Author Archives: admin

OGB Appreciation Day – Face Recognition in an easy way

How’s it going horse?

It’s Oracle Groundbreaker’s Appreciation Day!

Today it’s #ThanksOGB day and I decided to join the idea with a post about Face Recognition in an easy way using Oracle OCI Marketplace Nvidia image.

What is OGB Appreciation day?
OGB Appreciation day

Don’t forget to search for tweets with #ThanksOGB.

Oracle blog post my article about Face Recognition in 4 lines of code.

Today I want to show how you can create an Oracle OCI instance using the Marketplace and configure the environment to run this example easily with the Nvidia image.

Oracle OCI Marketplace

You can follow this link on how to create an image using the marketplace. Just change to use Nvidia image!

Once you have the instance running just ssh the image and run:

sudo apt install python3-pip
sudo apt-get install python3-setuptools
tar jxvf dlib-19.17.tar.bz2
cd dlib-19.17
sudo python3 install
sudo pip3 install face_recognition

Now you are ready to play with Face Recognition.

You can use my Python code to Streaming your Raspberry Pi camera feeds.

import io
import picamera
import logging
import socketserver
from threading import Condition
from http import server

<title>Raspberry Pi - Surveillance Camera</title>
<center><h1>Raspberry Pi - Surveillance Camera</h1></center>
<center><img src="stream.mjpg" width="640" height="480"></center>

class StreamingOutput(object):
    def __init__(self):
        self.frame = None
        self.buffer = io.BytesIO()
        self.condition = Condition()

    def write(self, buf):
        if buf.startswith(b'\xff\xd8'):
            # New frame, copy the existing buffer's content and notify all
            # clients it's available
            with self.condition:
                self.frame = self.buffer.getvalue()
        return self.buffer.write(buf)

class StreamingHandler(server.BaseHTTPRequestHandler):
    def do_GET(self):
        if self.path == '/':
            self.send_header('Location', '/index.html')
        elif self.path == '/index.html':
            content = PAGE.encode('utf-8')
            self.send_header('Content-Type', 'text/html')
            self.send_header('Content-Length', len(content))
        elif self.path == '/stream.mjpg':
            self.send_header('Age', 0)
            self.send_header('Cache-Control', 'no-cache, private')
            self.send_header('Pragma', 'no-cache')
            self.send_header('Content-Type', 'multipart/x-mixed-replace; boundary=FRAME')
                while True:
                    with output.condition:
                        frame = output.frame
                    self.send_header('Content-Type', 'image/jpeg')
                    self.send_header('Content-Length', len(frame))
            except Exception as e:
                    'Removed streaming client %s: %s',
                    self.client_address, str(e))

class StreamingServer(socketserver.ThreadingMixIn, server.HTTPServer):
    allow_reuse_address = True
    daemon_threads = True

with picamera.PiCamera(resolution='640x480', framerate=24) as camera:
    output = StreamingOutput()
    #Uncomment the next line to change your Pi's Camera rotation (in degrees)
    #camera.rotation = 90
    camera.start_recording(output, format='mjpeg')
        address = ('', 8000)
        server = StreamingServer(address, StreamingHandler)

Just change the code for instead of 0 use the streaming url, in the example here: http://ipaddress:8000/stream

video_capture = cv2.VideoCapture(0)

Happy face recognition.



What is Stream processing?

Hey you!

if you’re not familiar with Big Data or Data lake, I suggest you have a look at my previous post “What is Big Data?” and “What is data lake?” before.
This post is a collection of links, videos, tutorials, blogs and books that I found mixed with my opinion.

Table of contents

01. What is Stream processing?
02. Martin Kleppmann
03. Typical use cases
04. Pattern
05. Evaluation: Choose a Stream Processing Framework or a Product or Both?
06. Vertical vs. Horizontal Scaling
07. Streaming is better with SQL
08. Streaming Windows
09. Why Stream Processing
10. Final considerations
11. Book
12. Influence’s List
13. Links

Stream processing is becoming something like a “grand unifying paradigm” for data processing. Outgrowing its original space of real-time data processing, stream processing is becoming a technology that offers new approaches to data processing (including batch processing), real-time applications, and even distributed transactions.

1. What is Stream Processing?

Stream processing is the act of continuous incorporate new data to compute a result. In stream processing, the input data is unbounded and has no predetermined beginning or end. It simply forms a series of events that arrives at the stream processing system e.g. credit card transactions, clicks on a website, or sensor readings from internet of things devices.

Streaming is a data distribution technique where data producers write data records into an ordered data stream from which data consumers can read that data in the same order. Here is a simple data streaming diagram illustrating a data producer, a data stream and a data consumer

Each data streaming product makes a certain set of assumptions about the use cases and processing techniques to support. These assumptions leads to certain design choices, which affect what types of stream processing behaviour you can implement with them.

From wikipedia;

Stream processing is a computer programming paradigm, equivalent to dataflow programming, event stream processing, and reactive programming, that allows some applications to more easily exploit a limited form of parallel processing.
Stream Processing is a powerful technology that can scan huge volumes of data coming from sensors, credit card swipes, clickstreams and other inputs, and find actionable insights nearly instantaneously. For example, Stream Processing can detect a single fraudulent transaction in a stream containing millions of legitimate purchases, act as a recommendation engine to determine what ad or promotion to display for a particular customer while he or she is actually shopping or compute the optimal price for a car service ride in only a few seconds.

The term “Stream Processing” means that the data is coming into the processing engine as a continuous “stream” of events produced by some outside system or systems, and the processing engine works so fast that all decisions are made without stopping the data stream and storing the information first.

Streaming data and event-driven architectures are rising in popularity. The ideas have been around for a while, but technological and architectural advances have made into reality capabilities like stream processing and even function-based (aka “serverless”) computing. In many cases, the ability to act on data quickly is more valuable than a new method for batch-processing or historical data analysis.

I Googled about and I found;

Streaming is processing of data in motion.
Streaming is data that is continuously generated by different sources.
Streaming is the continuous high-speed transfer of large amounts of data from a source system to a target.
Programming paradigm that allows some applications to more easily exploit a limited form of parallel processing.

Streaming decouple data producers and data consumers from each other. When a data producer simply writes its data to a data stream, the producer does not need to know the consumers that read the data. Consumers can be added and removed independently of the producer. Consumers can also start and stop or pause and resume their consumption without the data producer needing to know about it. This decoupling simplifies the implementation of both data producers and consumers.

A data stream can be persistent, in which case it is sometimes referred to as a log or a journal. A persistent data stream has the advantage that the data in the stream can survive a shutdown of the data streaming service, so no data records are lost.
Persistent data streaming services can typically hold larger amounts of historic data than a data streaming service that only holds records in memory. Some data streaming services can even hold historic data all the way back to the first record written to the data stream. Others only hold e.g. a number of days of historic data.
In the cases where a persistent data stream holds the full history of records, consumers can replay all these records and recreate their internal state based on these records. In case a consumer discovers a bug in its own code, it can correct that code and replay the data stream to recreate its internal database.

2. Martin Kleppmann

Martin Kleppmann is the author of the book “Designing Data Intensive Applications”, and he has some nice papers/presentations;

Two ideas came from this;

  • All the BuzzWords are the same thing;
  • The concept of Streaming came from Database “Replication”;

1. BuzzWords

Some people call it stream processing. Others call it Event Sourcing or CQRS. Some even call it Complex Event Processing. Sometimes, such self-important buzzwords are just smoke and mirrors, invented by companies who want to sell you stuff. But sometimes, they contain a kernel of wisdom which can really help us design better systems.

The idea of structuring data as a stream of events is nothing new, and it is used in many different fields. Even though the underlying principles are often similar, the terminology is frequently inconsistent across different fields, which can be quite confusing. Although the jargon can be intimidating when you first encounter it, don’t let that put you off; many of the ideas are quite simple when you get down to the core.

But there’s some Differences.

In this article you can see some differences and similarities.

2. Replication

if we took that replication stream, and made it a first-class citizen in our data architecture? What if we changed our infrastructure so that the replication stream was not an implementation detail, but a key part of the public interface of the database? What if we turn the database inside out, take the implementation detail that was previously hidden, and make it a top-level concern? What would that look like?

3. Typical use cases

Stream Processing is rapidly gaining popularity and finding applications in various business domains. Found its first uses in the finance industry, as stock exchanges moved from floor-based trading to electronic trading. Today, it makes sense in almost every industry – anywhere where you generate stream data through human activities, machine data or sensors data. Assuming it takes off, the Internet of Things will increase volume, variety and velocity of data, leading to a dramatic increase in the applications for stream processing technologies.

Some use cases where stream processing can solve business problems include:

  • Network monitoring
  • Intelligence and surveillance
  • Risk management
  • E-commerce
  • Fraud detection
  • Smart order routing
  • Transaction cost analysis
  • Pricing and analytics
  • Market data management
  • Algorithmic trading
  • Data warehouse augmentation

Here is a short list of well-known, proven applications of Stream Processing:

  • Clickstream analytics can act as a recommendation engine providing actionable insights used to personalize offers, coupons and discounts, customize search results, and guide targeted advertisements — all of which help retailers enhance the online shopping experience, increase sales, and improve conversion rates.
  • Preventive maintenance allows equipment manufacturers and service providers to monitor quality of service, detect problems early, notify support teams, and prevent outages.
  • Fraud detection alerts banks and service providers of suspected frauds in time to stop bogus transactions and quickly notify affected accounts.
  • Emotions analytics can detect an unhappy customer and help customer service augment the response to prevent escalations before the customer’s unhappiness boils over into anger.
  • A dynamic pricing engine determines the price of a product on the fly based on factors such as current customer demand, product availability, and competitive prices in the area.

Common Usage Pattern for In-Stream Analytics

4. Pattern

Writing Streaming Applications requires very different thinking patterns from writing code with a language like Java. A better understanding of common patterns in Stream Processing will let us understand the domain better and build tools that handle those scenarios.

Pattern 1: Preprocessing

Preprocessing is often done as a projection from one data stream to the other or through filtering. Potential operations include

  • Filtering and removing some events
  • Reshaping a stream by removing, renaming, or adding new attributes to a stream
  • Splitting and combining attributes in a stream
  • Transforming attributes

For example, from a twitter data stream, we might choose to extract the fields: author, timestamp, location, and then filter them based on the location of the author.

Pattern 2: Alerts and Thresholds

This pattern detects a condition and generates alerts based on a condition. (e.g. Alarm on high temperature). These alerts can be based on a simple value or more complex conditions such as rate of increase etc.

For an example, in TFL (Transport for London) Demo video based on transit data from London, we trigger a speed alert when the bus has exceeded a given speed limit.

We can generate alerts for scenarios such as the server room temperature is continually increasing for the last 5 mins.

Pattern 3: Simple Counting and Counting with Windows

This pattern includes aggregate functions like Min, Max, Percentiles etc, and they can be counted without storing any data. (e.g. counting the number of failed transactions).

However, counts are often used with a time window attached to it. ( e.g. failure count last hour). There are many types of windows: sliding windows vs. batch (tumbling) windows and time vs. length windows. There are four main variations.

  • Time, Sliding window: keeps each event for the given time window, produce an output whenever a new event has added or removed.
  • Time, Batch window: also called tumbling windows, they only produce output at the end of the time window
  • Length, Sliding: same as the time, sliding window, but keeps a window of n events instead of selecting them by time.
  • Length, Batch window: same as the time, batch window, but keeps a window of n events instead of selecting them by time

There are special windows like decaying windows and unique windows.

Pattern 4: Joining Event Streams

The main idea behind this pattern is to match up multiple data streams and create a new event steam. For an example, let’s assume we play a football game with both the players and the ball having sensors that emit events with current location and acceleration. We can use “joins” to detect when a player has kicked the ball. To that end, we can join the ball location stream and the player stream on the condition that they are close to each other by one meter and the ball’s acceleration has increased by more than 55m/s^2.

Among other use cases are combining data from two sensors, and detecting the proximity of two vehicles. Please refer to Stream Processing 101: From SQL to Streaming SQL in 10 Minutes for more details.

Pattern 5: Data Correlation, Missing Events, and Erroneous Data

This pattern and the pattern four a has lot in common where here too we match up multiple streams. In addition, we also correlate the data within the same stream. This is because different data sensors can send events at different rates, and many use cases require this fundamental operator.

Following are some possible scenarios.

  • Matching up two data streams that send events at different speeds
  • Detecting a missing event in a data stream ( e.g. detect a customer request that has not been responded within 1 hour of its reception. )
  • Detecting erroneous data (e.g. Detect failed sensors using a set of sensors that monitor overlapping regions and using those redundant data to find erroneous sensors and removing their data from further processing)

Pattern 6: Interacting with Databases

Often we need to combine the real time data against the historical data stored in a disk. Following are a few examples.

  • When a transaction happened, look up the age using the customer ID from customer database to be used for Fraud detection (enrichment)
  • Checking a transaction against blacklists and whitelists in the database
  • Receive an input from the user (e.g. Daily discount amount may be updated in the database, and then the query will pick it automatically without human intervention.)

Pattern 7: Detecting Temporal Event Sequence Patterns

Using regular expressions with strings, we detect a pattern of characters from a sequence of characters. Similarly, given a sequence of events, we can write a regular expression to detect a temporal sequence of events arranged on time where each event or condition about the event is parallel to a character in a string in the above example.

A frequently cited example, although bit simplistic, is that a thief, having stolen a credit card, would try a smaller transaction to make sure it works and then do a large transaction. Here the small transaction followed by a large transaction is a temporal sequence of events arranged on time and can be detected using a regular expression written on top of an event sequence.

Such temporal sequence patterns are very powerful. For example, the following video shows a real time analytics done using the data collected from a real football game. This was the dataset taken from DEBS 2013 Grand Challenge.

In the video, we used patterns on event sequence to detect the ball possession, the time period a specific player controlled the ball. A player possessed the ball from the time he hits the ball until someone else hits the ball. This condition can be written as a regular expression: a hit by me, followed by any number of hits by me, followed by a hit by someone else. (We already discussed how to detect the hits on the ball in Pattern 4: Joins).

Pattern 8: Tracking

The eighth pattern tracks something over space and time and detects given conditions.
Following are few examples

  • Tracking a fleet of vehicles, making sure that they adhere to speed limits, routes, and geo-fences.
  • Tracking wildlife, making sure they are alive (they will not move if they are dead) and making sure they will not go out of the reservation.
  • Tracking airline luggage and making sure they are not been sent to wrong destinations
  • Tracking a logistic network and figure out bottlenecks and unexpected conditions.

For example, TFL Demo we discussed under pattern 2 shows an application that tracks and monitors London buses using the open data feeds exposed by TFL(Transport for London).

Pattern 9: Detecting Trends

We often encounter time series data. Detecting patterns from time series data and bringing them into operator attention are common use cases.
Following are some of the examples of tends.

  • Rise, Fall
  • Turn (switch from a rise to a fall)
  • Outliers
  • Complex trends like triple bottom etc.

These trends are useful in a wide variety of use cases such as

  • Stock markets and Algorithmic trading
  • Enforcing SLA (Service Level Agreement), Auto Scaling, and Load Balancing
  • Predictive maintenance ( e.g. guessing the Hard Disk will fill within next week)

Pattern 10: Running the same Query in Batch and Realtime Pipelines

This pattern runs the same query in both Relatime and batch pipeline. It is often used to fill the gap left in the data due to batch processing. For example, if batch processing takes 15 minutes, results would lack the data for the last 15 minutes.

The idea of this pattern, which is sometimes called “Lambda Architecture” is to use real time analytics to fill the gap. Jay Kreps’s article “Questioning the Lambda Architecture” discusses this pattern in detail.

Pattern 11: Detecting and switching to Detailed Analysis

The main idea of the pattern is to detect a condition that suggests some anomaly, and further analyze it using historical data. This pattern is used with the use cases where we cannot analyze all the data with full detail. Instead, we analyze anomalous cases in full detail. Following are a few examples.

    Use basic rules to detect Fraud (e.g. large transaction), then pull out all transactions done against that credit card for a larger time period (e.g. 3 months data) from a batch pipeline and run a detailed analysis
  • While monitoring weather, detect conditions like high temperature or low pressure in a given region and then start a high resolution localized forecast on that region.
  • Detect good customers, for example through the expenditure of more than $1000 within a month, and then run a detailed model to decide the potential of offering a deal.

Pattern 12: Using a Model

The idea is to train a model (often a Machine Learning model), and then use it with the Realtime pipeline to make decisions. For example, you can build a model using R, export it as PMML (Predictive Model Markup Language) and use it within your realtime pipeline.

Among examples is Fraud Detections, Segmentation, Predict next value, Predict Churn. Also see InfoQ article, Machine Learning Techniques for Predictive Maintenance, for a detailed example of this pattern.

Pattern 13: Online Control

There are many use cases where we need to control something online. The classical use cases are the autopilot, self-driving, and robotics. These would involve problems like current situation awareness, predicting the next value(s), and deciding on corrective actions.

You can implement most of these use cases with a Stream Processor that supports a Streaming SQL language.

This pattern list came from (9th ACM International Conference on Distributed Event-Based Systems), describing a set of real time analytics patterns.

You can find details about pattern implementations and source code from here.

Monal Daxini presents a blueprint for streaming data architectures and a review of desirable features of a streaming engine. He also talks about streaming application patterns and anti-patterns, and use cases and concrete examples using Apache Flink.
Patterns of Streaming Applications

5. Evaluation: Choose a Stream Processing Framework or a Product or Both?

There are many different data streaming products, and it can be hard to know where to start studying them, and which products do what etc.

The typical evaluation process (long list, short list, proof of concept) is obligatory before making a decision.

  • A stream processing programming language for streaming analytics
  • Visual development and debugging instead of coding
  • Real-time analytics
  • Monitoring and alerts
  • Support for fault tolerance, and highly optimized performance
  • Product maturity
  • In the case of TIBCO, a live data mart and operational command and control center for business users
  • Out-of-the-box connectivity to plenty of streaming data sources
  • Commercial support
  • Professional services and training.

Think about which of the above features you need for your project. In addition, you have to evaluate the costs of using a framework against productivity, reduced effort and time-to-market using a product before making your choice.

Besides evaluating the core features of stream processing products, you also have to check integration with other products. Can a product work together with messaging, Enterprise Service Bus (ESB), Master Data Management (MDM), in-memory stores, etc. in a loosely coupled, but highly integrated way? If not, there will be a lot of integration time and high costs.

6. Vertical vs. Horizontal Scaling

Vertical scaling means running your data streaming storage and processors on a more powerful computer. Vertical scaling is also sometimes referred to as scaling up. You scale up the size and speed of its disk, memory, speed of CPUs, possibly CPU cores too, graphics cards etc.

Horizontal scaling means distributing the workload among multiple computers. Thus, the data in the data stream is distributed among multiple computers, and the applications processing the data streams are too (or at least they can be). Horizontal scaling is also sometimes referred to as scaling out. You scale out from a single computer to multiple computers.

Distributing the messages of a data stream onto multiple computers is also referred to as partitioning the data stream.

1. Round Robin Partitioning

Round robin data stream partitioning is the simplest way to partition the messages of a data stream across multiple computers. The round robin partitioning method simply distributes the messages evenly and sequentially among the computers. In other words, the first message is stored on the first computer, the second message on the second computer etc. When all computers have received a message from the stream, the round robin method starts from the first computer again.

2. Key Based Partitioning

Key based partitioning distributes the message across different computers based on a certain key value read from each message. Commonly the identifying id (e.g. primary key) is used as key to distribute the messages. Typically, a hash value is calculated from each key value, and that hash value is then used to map the message to one of the computers in the cluster.

Stream Processing and DWH

A DWH is a great tool to store and analyze structured data. You can store terabytes of data and get answers to your queries about historical data within seconds. DWH products such as Teradata or HP Vertica were built for this use case. However the ETL processes often take too long. Business wants to query up-to-date information instead of using an approach where you may only get information about what happened yesterday. This is where stream processing comes in and feeds all new data into the DWH immediately.

Stream Processing and Hadoop

A big data architecture contains stream processing for real-time analytics and Hadoop for storing all kinds of data and long-running computations.

Hadoop initially started with MapReduce, which offers batch processing where queries take hours, minutes or at best seconds. This is and will be great for complex transformations and computations of big data volumes. However, it is not so good for ad hoc data exploration and real-time analytics. Multiple vendors have though made improvements and added capabilities to Hadoop that make it capable of being more than just a batch framework.

DWH, Hadoop and stream processing complement each other very well. Therefore, the integration layer is even more important in the big data era, because you have to combine more and more different sinks and sources.

Since 2016, a new idea called Streaming SQL has emerged. We call a language that enables users to write SQL like queries to query streaming data as a “Streaming SQL” language. Almost all Stream Processors now support Streaming SQL.

7. Streaming is better with SQL

Let’s assume that you picked a stream processor, implemented some use cases, and it’s working. Now you sit down to savor the win. However, given that you can simply write SQL or something like SQL when doing batch processing, why should you have to write all this code? Shouldn’t you be able to do streaming with SQL? The answer is yes, you should. Such streaming SQL exists. Again there are many offerings. Unfortunately, unlike SQL, there is no standard streaming SQL syntax. There are many favors, which follow SQL but have variations.

SQL is a powerful language for querying structured data. It is designed as a set of independent operators: projection, filter, joins, and grouping, which can be recombined to create very powerful queries.

Following are some advantages of streaming SQL languages:

  • It’s easy to follow and learn for the many people who know SQL.
  • It’s expressive, short, sweet and fast!!
  • It defines core operations that cover 90% of problems.
  • Streaming SQL language experts can dig in when they like by writing extensions!
  • A query engine can better optimize the executions with a streaming SQL model. Most optimizations are already studied under SQL, and there is much we can simply borrow from database optimizations.

Let us walk through a few of the key operators. Just as SQL can cover most data queries on data stored in a disk, streaming SQL can cover most of the queries on streaming data. Without streaming SQL, programmers would have to hand code each operator, which is very complicated and hard work.

Concepts in SQL, such as “group by” and “having” clauses, usually work similarly with streaming SQL languages.

Streaming SQL has two additional concepts not covered by SQL: windows and joins, which handle the complexities of streaming. Let’s understand each of them.

8. Streaming Windows

Although batch can be handled as a special case of stream processing, analyzing never-ending streaming data often requires a shift in the mindset and comes with its own terminology (for example, “windowing” and “at-least-once”/”exactly-once” processing). This shift and the new terminology can be quite confusing for people being new to the space of stream processing.

Consider the example of a traffic sensor that counts every 15 seconds the number of vehicles passing a certain location. The resulting stream could look like:

If you would like to know, how many vehicles passed that location, you would simply sum the individual counts. However, the nature of a sensor stream is that it continuously produces data. Such a stream never ends and it is not possible to compute a final sum that can be returned. Instead, it is possible to compute rolling sums, i.e., return for each input event an updated sum record. This would yield a new stream of partial sums.

However, a stream of partial sums might not be what we are looking for, because it constantly updates the count and even more important, some information such as variation over time is lost. Hence, we might want to rephrase our question and ask for the number of cars that pass the location every minute. This requires us to group the elements of the stream into finite sets, each set corresponding to sixty seconds. This operation is called a tumbling windows operation.

Tumbling windows discretize a stream into non-overlapping windows. For certain applications it is important that windows are not disjunct because an application might require smoothed aggregates. For example, we can compute every thirty seconds the number of cars passed in the last minute. Such windows are called sliding windows.

This is because each element of a stream must be processed by the same window operator that decides which windows the element should be added to. For many applications, a data stream needs to be grouped into multiple logical streams on each of which a window operator can be applied. Think for example about a stream of vehicle counts from multiple traffic sensors (instead of only one sensor as in our previous example), where each sensor monitors a different location. By grouping the stream by sensor id, we can compute windowed traffic statistics for each location in parallel.
The following figure shows tumbling windows that collect two elements over a stream of (sensorId, count) pair elements.

Generally speaking, a window defines a finite set of elements on an unbounded stream. This set can be based on time (as in our previous examples), element counts, a combination of counts and time, or some custom logic to assign elements to windows.

Streaming Joins

If we want to handle data from multiple tables, we use the JOIN operator in SQL. Similarly, if you want to handle data from multiple streams, there are two options. First is to join the two and create one stream while the second is to write patterns across multiple streams.

9. Why Stream Processing

10. Final considerations

We have entered an era where competitive advantage comes from analyzing, understanding, and responding to an organization’s data. When doing this, time is of the essence, and speed will decide the winners and losers.

Stream processing is required when data has to be processed fast and / or continuously, i.e. reactions have to be computed and initiated in real time. This requirement is coming more and more into every vertical. Many different frameworks and products are available on the market already.

Many use cases need fast, real-time decisions. Although it is possible to implement them using databases or batch processing, these technologies quickly introduce complexities because there is a fundamental impedance mismatch between the use cases and the tools employed. In contrast, streaming provides a much more natural model to think about, capture, and implement those real-time streaming use cases. Streaming SQL provides a simple yet powerful language to program streaming use cases.

The reality is that the value of most data degrades with time. It’s interesting to know that yesterday there was a traffic jam, or 10 fraud incidents, or 10 people who had heart attacks. From that knowledge, we can learn how to mitigate or prevent those incidents in the future. However, it is much better if we can gain those insights at the time they are occurring so that we can intervene and manage the situation.

The most popular Stream processing framework is Kafka. You can check my previous post here

What is Kafka?

11. Book

Designing Data-Intensive Applications

Stream processing book bundle

Streaming Systems

Event Streams in action

12. Influencers List


13. Links

The Log: What every software engineer should know about real-time data’s unifying abstraction

Oracle understanding stream analytics

Class 101

Class 102

Stream processing myths debunked – Six Common Streaming Misconceptions
Myth 1: There’s no streaming without batch (the Lambda Architecture)
Myth 2: Latency and Throughput: Choose One
Myth 3: Micro-batching means better throughput
Myth 4: Exactly once? Completely impossible.
Myth 5: Streaming only applies to “real-time”
Myth 6: So what? Streaming is too hard anyway.

The data processing evolution a potted history

Choosing a stream processor is challenging because there are many options to choose from and the best choice depends on end-user use cases.

How to choose stream processor

Streaming first architecture

Migrating to an event driven system

More details about Stream and SQL

Predict Conference 2019

What’s the crack jack?

Yesterday I went to Predict Conference, and I need to say thanks to Oracle for providing this experience for me.


My first time at Predict Conference, my first impression: it is a nice conference to get what are the trending topics in the area of Data in Ireland.
With nearly 1000 attendees, more than 30 speakers, over 2 stages.
Women are more and more present in tech.

But what is Predict Conference?

The Predict Conference is unique in its approach to combining business, data science and technology under one roof.

Predict is a holistic and continuous experience. It combines great talks, inspiration, face-to-face serendipity, experience zone, and hands-on data modeling innovation to equip you for decision making in the data age.
Predict is Europe’s leading data conference. It is designed to bring together thought leaders and innovators in the fields of data science, predictive analytics, artificial intelligence and technology.

Tickets include:
– Full access to the main event on the 1st of October 2019: all talks, access to the Predict experience area, all-day coffee, lunch and evening drinks reception.

How many years has it been going?

Creme Global started Predict in 2015; it is now in its 5th year and has gone from strength to strength.
What exciting things can people look forward to?
The major themes at Predict this year where Health & Life Sciences, AI and Machine Learning along with Technology and Society (Sustainability/ Government/ Cities/ Privacy);

It’s not a very deep dive tech conference

The Predict Conference is just about talks around data topics, but it isn’t the most techie one for sure. The idea of this conference was to bring the tech people and industries together — and this objective is being realized year after year with great success.

From video analysis to APIs and IOT

The range of topics was very broad: but for me, everything was around 3 main topics; Video analizes, text analysis and IOT.

The talks were divided into some main categories;
Sustainability & Cities;
Customer first;

I create my list of talks that I had the opportunity to see there. This is not an exhaustive list!

From Cups to Consciousness: The Roadmap to AGI – Ben Duffy
He talked about AGI, Artificial general intelligence that is the intelligence of a machine that has the capacity to understand or learn any intellectual task that a human being can. It is a primary goal of some artificial intelligence research and a common topic in science fiction and future studies, and also open the floor for debate about what the development of AI and robotics will look like over the next few years.

Onboard Artificial Intelligence: train, deploy and use Deep Learning on an edge device (Raspberry Pi) – Constant Bridon
He showed a live demo of a model designed to recognised car drawings via the camera of a Raspberry Pi. He also talked about an IOT car competition in the Us where the car run using the feeds of the camera.
There’s a nice one in UK called FormulaPI.

The history of Creative AI – Eric Risser
He talked about artificial intelligence and computer graphics to lead both the vision and the core technology.

There’s a nice colleague video about showing how do do it a demo using Oracle Cloud and GPU

Top 3 things I’ve learned in 3 decades of Data Science – Dr. John F Elder
He talked about Ensembles, we can’t do it alone, we need alternative perspectives, Target Shuffling, some of our success is luck, simulation can find out how much and Leaks from the future, we can’t blindly trust results, if it looks good to good to be true, it is.

Uncertainty and Interpretability – Kevin Kuo

Low-Latency Model Prediction with Video
He talked about TensorFlow and buy the way there’s a 2.0 version now.

The Fast Track to AI with Serverless – Peter Elger

AI Live: No Experience Required
He showed one Wolfram notebook.

But the main topic, in my opinion, was ML & DL buzzwords

A lot of presentation was trying to explain that Machine Learning and Deep Learning are not the same things, and that there’s a lot of other buzzwords around.


A stands area with some nice content. Experience Zone.

Wolfram – they are showing case the computing program Wolfram Mathematica.

Equinix – specializes in internet connection and data centers

Rstudio – they are showing case the IDE and language. I had a nice talk with the guy there about R is not dying and is Python that becoming more strong.

Coder Dojo –

CeADAR – they had a nice tic tac toe game.


Led Race – Another DublinMaker

Hey you!

Last week 20th July was the Dublin Maker 2019.

So, let’s start with an explanation about what is Dublin Maker.

Dublin Maker is a free to attend, community run event, which was held on Saturday, July 20th, 2019 in Merrion Square. Dublin Maker takes the form of a “show and tells” experience where inventors/makers sourced through an open call, to have an opportunity to showcase their creations in a carnival atmosphere. It is a family friendly showcase of invention, creativity and resourcefulness, and a celebration of the maker movement. It’s a place where people show what they are making and share what they are learning.

If you have science interested kids, or you’re a kid yourself, this is a great event with lots of interesting open people and things that at least some of them you’ll probably not get the chance to see again.

I have been attending the event for a long time now, and today I want to explain the project that I presented this year. The Open Led Race. Back in March I was on holidays and I saw one tweet from Arduino about the project and I said to myself, it’s exactly what I’m going to show on Dublin Maker.

My first concern was about using an arcade button, because the idea is to let a lot of kids to play, I was afraid that they would break the button fast and I need to replace fast as well because a lot of people came to see the project.
After few days looking at the instructions and the components I had the idea to change the arcade button to MakeyMakey and because of that I decided to create the idea using a Raspberry PI to simplify the idea.

Apparent I’m the first one to do it.

The project is really simple. I’m using the WS2813 led strip and using the API that I found on the internet.

Python library wrapping for the rpi-ws281x library

You can check my GitHub to see the full code.

It’s basically a Python code that runs on the Raspberry Pi that controls the Led Strip and the MakeyMakey that I used to simulate one keyboard. Every click I move the led 3 positions forward.
The makeymakey part is just one aluminium foil and play-doh.

I used the GPIO 10 (pin 19) and 18 (pin 12) for LED_PIN and LED_DMA and GPIO 9 (pin 6) for Ground.

I created a simple version of the Led Race, but there are lots of space for improvement.

I came with some ideas that one day I’ll implement;

  • Add a monitor where I can show a timer and the best lap timer;
  • I can show a speedometer or something like the number of push per second;
  • I can display the best lap overall;
  • four players put for cars at the same time;

Other ideas are some things that I saw on the Open Led Races Arduino web site and the comments, like;

  • Add velocity;
  • Add some physics when the car goes up, more push is needed, or increase the speed when going down;
  • Add a second Led Strip and then the car can go left and right and they can leave some kind of weapons on the track, and the car gets stuck there if hits for fill push;

I want to add here a big thanks to Elaine Akemi who helped me with the project. She is also my official partner of Hackathons and events, and she was with me in the last two Dublin Maker editions.

Hadoop Ecosystem & Hadoop Distributions

Alright boss?

The complexity of Hadoop ecosystem comes from having many tools that can be used, and not from using Hadoop itself.

The objective of this Apache Hadoop ecosystem components tutorial is to have an overview of what are the different components of Hadoop ecosystem that make Hadoop so powerful, to give you a nice overview of some Hadoop related products and about somel Hadoop distributions in the market. I did some research and this is what I know or found and probably is not a exhaustive list!

if you’re not familiar with Big Data or Data lake, I suggest you have a look on my previous post “What is Big Data?” and “What is data lake?” before.
This post is a collection of links, videos, tutorials, blogs and books that I found mixed with my opinion.

Hadoop Ecosystem

What does Hadoop Ecosystem mean?

Hadoop Ecosystem is neither a programming language nor a service, it is a platform or framework which solves big data problems. You can consider it as a suite which encompasses a number of services (ingesting, storing, analyzing and maintaining) inside it. The Hadoop ecosystem includes both official Apache open source projects and a wide range of commercial tools and solutions. Most of the solutions available in the Hadoop ecosystem are intended to supplement one or two of Hadoop’s four core elements (HDFS, MapReduce, YARN, and Common). However, the commercially available framework solutions provide more comprehensive functionality.

The Hadoop Ecosystem Table


Hive is data warehousing software that addresses how data is structured and queried in distributed Hadoop clusters. Hive is also a popular development environment that is used to write queries for data in the Hadoop environment. It provides tools for ETL operations and brings some SQL-like capabilities to the environment. Hive is a declarative language that is used to develop applications for the Hadoop environment, however it does not support real-time queries.

Pig is a procedural language for developing parallel processing applications for large data sets in the Hadoop environment. Pig is an alternative to Java programming for MapReduce, and automatically generates MapReduce functions. Pig includes Pig Latin, which is a scripting language. Pig translates Pig Latin scripts into MapReduce, which can then run on YARN and process data in the HDFS cluster. Pig is popular because it automates some of the complexity in MapReduce development.

HBase is a scalable, distributed, NoSQL database that sits atop the HFDS. It was designed to store structured data in tables that could have billions of rows and millions of columns. It has been deployed to power historical searches through large data sets, especially when the desired data is contained within a large amount of unimportant or irrelevant data (also known as sparse data sets). It is also an underlying technology behind several large messaging applications, including Facebook’s.

Oozie is the workflow scheduler that was developed as part of the Apache Hadoop project. It manages how workflows start and execute, and also controls the execution path. Oozie is a server-based Java web application that uses workflow definitions written in hPDL, which is an XML Process Definition Language similar to JBOSS JBPM jPDL. Oozie only supports specific workflow types, so other workload schedulers are commonly used instead of or in addition to Oozie in Hadoop environments.

Think of Sqoop as a front-end loader for big data. Sqoop is a command-line interface that facilitates moving bulk data from Hadoop into relational databases and other structured data stores. Using Sqoop replaces the need to develop scripts to export and import data. One common use case is to move data from an enterprise data warehouse to a Hadoop cluster for ETL processing. Performing ETL on the commodity Hadoop cluster is resource efficient, while Sqoop provides a practical transfer method.

It is a table and storage management layer for Hadoop. HCatalog supports different components available in Hadoop ecosystems like MapReduce, Hive, and Pig to easily read and write data from the cluster. HCatalog is a key component of Hive that enables the user to store their data in any format and structure.
By default, HCatalog supports RCFile, CSV, JSON, sequenceFile and ORC file formats.

Acro is a part of Hadoop ecosystem and is a most popular Data serialization system. Avro is an open source project that provides data serialization and data exchange services for Hadoop. These services can be used together or independently. Big data can exchange programs written in different languages using Avro.

It is a software framework for scalable cross-language services development. Thrift is an interface definition language for RPC(Remote procedure call) communication. Hadoop does a lot of RPC calls so there is a possibility of using Hadoop Ecosystem componet Apache Thrift for performance or other reasons.

The main purpose of the Hadoop Ecosystem Component is large-scale data processing including structured and semi-structured data. It is a low latency distributed query engine that is designed to scale to several thousands of nodes and query petabytes of data. The drill is the first distributed SQL query engine that has a schema-free model.

Mahout is open source framework for creating scalable machine learning algorithm and data mining library. Once data is stored in Hadoop HDFS, mahout provides the data science tools to automatically find meaningful patterns in those big data sets.

Flume efficiently collects, aggregate and moves a large amount of data from its origin and sending it back to HDFS. It is fault tolerant and reliable mechanism. This Hadoop Ecosystem component allows the data flow from the source into Hadoop environment. It uses a simple extensible data model that allows for the online analytic application. Using Flume, we can get the data from multiple servers immediately into hadoop.

Ambari, another Hadop ecosystem component, is a management platform for provisioning, managing, monitoring and securing apache Hadoop cluster. Hadoop management gets simpler as Ambari provide consistent, secure platform for operational control.

Apache Zookeeper is a centralized service and a Hadoop Ecosystem component for maintaining configuration information, naming, providing distributed synchronization, and providing group services. Zookeeper manages and coordinates a large cluster of machines.

Apache Lucene is a full-text search engine which can be used from various programming languages, is a free and open-source information retrieval software library, originally written completely in Java by Doug Cutting. It is supported by the Apache Software Foundation and is released under the Apache Software License.

Solr is an open-source enterprise-search platform, written in Java, from the Apache Lucene project. Its major features include full-text search, hit highlighting, faceted search, real-time indexing, dynamic clustering, database integration, NoSQL features and rich document handling.

Apache Phoenix is an open source, massively parallel, relational database engine supporting OLTP for Hadoop using Apache HBase as its backing store.

Presto is a high performance, distributed SQL query engine for big data. Its architecture allows users to query a variety of data sources such as Hadoop, AWS S3, Alluxio, MySQL, Cassandra, Kafka, and MongoDB. One can even query data from multiple data sources within a single query.

Apache Zeppelin is a multi-purposed web-based notebook which brings data ingestion, data exploration, visualization, sharing and collaboration features to Hadoop and Spark.

Apache Storm is a distributed stream processing computation framework written predominantly in the Clojure programming language. Originally created by Nathan Marz and team at BackType, the project was open sourced after being acquired by Twitter.

Apache Flink is an open-source stream-processing framework developed by the Apache Software Foundation. The core of Apache Flink is a distributed streaming data-flow engine written in Java and Scala. Flink executes arbitrary dataflow programs in a data-parallel and pipelined manner.

Apache Samza is an open-source near-realtime, asynchronous computational framework for stream processing developed by the Apache Software Foundation in Scala and Java.

Apache Arrow is a cross-language development platform for in-memory data. It specifies a standardized language-independent columnar memory format for flat and hierarchical data, organized for efficient analytic operations on modern hardware.

Airflow is a platform to programmatically author, schedule and monitor workflows. Use airflow to author workflows as directed acyclic graphs (DAGs) of tasks. The airflow scheduler executes your tasks on an array of workers while following the specified dependencies.

BlazingSQL is a distributed GPU-accelerated SQL engine with data lake integration, where data lakes are huge quantities of raw data that are stored in a flat architecture. It is ACID-compliant. BlazingSQL targets ETL workloads and aims to perform efficient read IO and OLAP querying. BlazingDB refers to the company and BlazingSQL refers to the product.

A data collection system for managing large distributed systems.

The open source, native analytic database for Apache Hadoop. Impala is shipped by Cloudera, MapR, Oracle, and Amazon.

Get started with Hadoop: From evaluation to your first production cluster

Hadoop distributions

When talk about HAdoop distribution the top 3 most famous are Cloudera, Hortonworks and MapR.

September 2018 Cloudera and Hortonworks announcing a merge to be completed summer 2019.

Cloudera and Hortonworks Announce Merger to Create World’s Leading Next Generation Data Platform and Deliver Industry’s First Enterprise Data Cloud

Cloudera and Hortonworks Complete Planned Merger

  • Cloudera offers the highest performance and lowest cost platform for using data to drive better business outcomes. Cloudera has a track record of bringing new open source solutions into its platform (such as Apache Spark, Apache HBase, and Apache Parquet) that are eventually adopted by the community at large. Cloudera Navigator provides everything your organization needs to keep sensitive data safe and secure while still meeting compliance requirements. Cloudera Manager is the easiest way to administer Hadoop in any environment, with advanced features like intelligent configuration defaults, customized monitoring, and robust troubleshooting. Cloudera delivers the modern data management and analytics…
  • Hortonworks Sandbox is a personal, portable Apache Hadoop environment that comes with dozens of interactive Hadoop and it’s ecosystem tutorials and the most exciting developments from the latest HDP distribution. Hortonworks Sandbox provides performance gains up to 10 times for applications that store large datasets such as state management, through a revamped Spark Streaming state tracking API. It provides seamless Data Access to achieve higher performance with Spark. Also provides dynamic Executor Allocation to utilize cluster resources efficiently through Dynamic Executor Allocation functionality that automatically expands and shrinks resources based on utilization. Hortonworks Sandbox
  • MapR Converged Data Platform integrates the power of Hadoop and Spark with global event streaming, real-time database capabilities, and enterprise storage for developing and running innovative data applications. Modules include MapR-FS, MapR-DB, and MapR Streams. Its enterprise- friendly design provides a familiar set of file and data management services, including a global namespace, high availability, data protection, self-healing clusters, access control, real-time performance, secure multi-tenancy, and management and monitoring. MapR tests and integrates open source ecosystem projects such as Hive, Pig, Apache HBase and Mahout, among others. MapR Community

Commercial Hadoop Vendors

1) Amazon Elastic MapReduce
2) Microsoft Azure’s HDInsight – Cloud based Hadoop Distribution
3) IBM Open Platform
4) Pivotal Big Data Suite
5) Datameer Professional
6) Datastax Enterprise Analytics
7) Dell – Cloudera Apache Hadoop Solution.
8) Oracle

Top Hadoop Appliances

Hadoop Appliances providers offer hardware optimised for Apache Hadoop or enterprise versions .

Dell provides PowerEdge servers, Cloudera Enterprise Basic Edition and Dell Professional Services, Dell PowerEdge servers with Intel Xeon processors, Dell Networking and Cloudera Enterprise and Dell In-Memory Appliance for Cloudera Enterprise with Apache Spark.

EMC provides Greenplum HD and Greenplum MR. EMC provides Pivotal HD, which is an Apache Hadoop distribution that natively integrates EMC Greenplum massively parallel processing (MPP) database technology with the Apache Hadoop framework.

Teradata Appliance for Hadoop provides optimized hardware, flexible configurations, high-speed connectors, enhanced software usability features, proactive systems monitoring, intuitive management portals, continuous availability, and linear scalability.

HP AppSystem for Apache Hadoop is an enterprise ready Apache Hadoop platform and provides RHEL v6.1, Cloudera Enterprise Core – the market leading Apache Hadoop software, HP Insight CMU v7.0 and a sandbox that includes HP Vertica Community Edition v6.1

NetApp Open Solution for Hadoop provides a ready to deploy, enterprise class infrastructure for the Hadoop platform to control and gain insights from big data.

Oracle Big Data Appliance X6-2 Starter Rack contains six Oracle Sun x86 servers within a full-sized rack with redundant Infiniband switches and power distribution units. Includes all Cloudera Enterprise Technology software including Cloudera CDH, Cloudera Manager, and Cloudera RTQ (Impala).

Top Hadoop Managed Services

Amazon EMR
Amazon EMR simplifies big data processing, providing a managed Hadoop framework that makes it easy, fast, and cost effective way to distribute and process vast amounts data across dynamically scalable Amazon EC2 instances.

Microsoft HDInisght
HDInsight is a managed Apache Hadoop, Spark, R, HBase, and Storm cloud service made easy. It provides a Data Lake service, Scale to petabytes on demand, Crunch all data structured, semi structured, unstructured and Develop in Java, .NET, and more. Provides Apache Hadoop, Spark, and R clusters in the cloud.

Google Cloud Platform
Google offers Apache Spark and Apache Hadoop clusters easily on Google Cloud Platform.

Qubole Data Service (QDS) offers Hadoop as a Service and is a cloud computing solution that makes medium and large-scale data processing accessible, easy, fast and in

IBM BigInsights
IBM BigInsights on Cloud provides Hadoop-as-a-service on IBM’s SoftLayer global cloud infrastructure. It offers the performance and security of an on-premises deployment.

Teradata Cloud for Hadoop
Teradata Cloud for Hadoop includes Teradata developed software components that make Hadoop ready for the enterprise: high availability, performance, scalability, monitoring, manageability, data transformation, data security, and a full range of tools and utilities.

Altiscale Data Cloud
Altiscale Data Cloud is a fully managed Big Data platform, delivering instant access to production ready Apache Hadoop and Apache Spark on the world’s best Big Data infrastructure.

Rackspace Apache
Rackspace Apache Hadoop distribution includes common tools like MapReduce, HDFS, Pig, Hive, YARN, and Tez. Rackspace provide root access to the application itself, allowing users to interact directly with the core platform.

Oracle offers a Cloudera solution on the top of the Oracle cloud infrastructure.


Hadoop Ecosystem and Their Components – A Complete Tutorial

Big Data & Data Lake a complete overview

What’s the crack jack?

If you ever wanted to know what is Big Data and not what you think Big Data is or If you ever wanted to know what is Data Lake and not what you think Data Lake is, you should check this out.

I just finished a series of blog post where I did an overview in Big Data, Data Lake, Hadoop, Apache Spark and Apache Kafka.

The idea here is a complete post with a good overview and a good start point to discover these areas and technologies.

What is Big Data?

What is Data Lake?

What is Hadoop?

What is Apache Spark?

What is Apache Kafka?

All post are based on a collection of links, videos, tutorials, blogs and books that I found mixed with my opinion.

There content to spend two hours reading, so good studies!

Thank you for taking the time to read this post.

What is Kafka?

How’s the form?

if you’re not familiar with Big Data or Data lake, I suggest you have a look on my previous post “What is Big Data?” and “What is data lake?” before.
This post is a collection of links, videos, tutorials, blogs and books that I found mixed with my opinion.

Table of contents

1. What is Kafka?
2. Architecture
3. History
4. Courses
5. Books
6. Influencers List
7. Link

1. What is Kafka?

In simple terms, Kafka is a messaging system that is designed to be fast, scalable, and durable. It is an open-source stream processing platform. Kafka is a distributed publish-subscribe messaging system that maintains feeds of messages in partitioned and replicated topics.

Wikipedia definition: Apache Kafka is an open-source stream-processing software platform developed by LinkedIn and donated to the Apache Software Foundation, written in Scala and Java. The project aims to provide a unified, high-throughput, low-latency platform for handling real-time data feeds. Its storage layer is essentially a “massively scalable pub/sub message queue designed as a distributed transaction log, making it highly valuable for enterprise infrastructures to process streaming data. Additionally, Kafka connects to external systems (for data import/export) via Kafka Connect and provides Kafka Streams, a Java stream processing library.

I Googled about and I found …
Kafka is designed for distributed high throughput systems. Kafka tends to work very well as a replacement for a more traditional message broker. In comparison to other messaging systems, Kafka has better throughput, built-in partitioning, replication and inherent fault-tolerance, which makes it a good fit for large-scale message processing applications.

Other …
Apache Kafka is a distributed publish-subscribe messaging system and a robust queue that can handle a high volume of data and enables you to pass messages from one end-point to another. Kafka is suitable for both offline and online message consumption. Kafka messages are persisted on the disk and replicated within the cluster to prevent data loss. Kafka is built on top of the ZooKeeper synchronization service. It integrates very well with Apache Storm and Spark for real-time streaming data analysis.


Producers produce messages to a topic of their choice. It is possible to attach a key to each message, in which case the producer guarantees that all messages with the same key will arrive to the same partition.


Consumers read the messages of a set of partitions of a topic of their choice at their own pace. If the consumer is part of a consumer group, i.e. a group of consumers subscribed to the same topic, they can commit their offset. This can be important if you want to consume a topic in parallel with different consumers.

Topics and Logs

A topic is a feed name or category to which records are published. Topics in Kafka are always multi-subscriber — that is, a topic can have zero, one, or many consumers that subscribe to the data written to it. For each topic, the Kafka cluster maintains a partition log that looks like this:

Topics are logs that receive data from the producers and store them across their partitions. Producers always write new messages at the end of the log.


A topic may have many partitions so that it can handle an arbitrary amount of data. In the above diagram, the topic is configured into three partitions (partition{0,1,2}). Partition 0 has 13 offsets, Partition 1 has 10 offsets, and Partition 2 has 13 offsets.

Partition Offset

Each partitioned message has a unique sequence ID called an offset. For example, in Partition 1, the offset is marked from 0 to 9. The offset is the position in the log where the consumer last consumed or read a message.


The partitions of the log are distributed over the servers in the Kafka cluster with each server handling data and requests for a share of the partitions. Each partition is replicated across a configurable number of servers for fault tolerance.

Each partition has one server which acts as the “leader” and zero or more servers which act as “followers”. The leader handles all read and write requests for the partition while the followers passively replicate the leader. If the leader fails, one of the followers will automatically become the new leader. Each server acts as a leader for some of its partitions and a follower for others so load is well balanced within the cluster.


Kafka MirrorMaker provides geo-replication support for your clusters. With MirrorMaker, messages are replicated across multiple datacenters or cloud regions. You can use this in active/passive scenarios for backup and recovery; or in active/active scenarios to place data closer to your users, or support data locality requirements.


Replicas are nothing but backups of a partition. If the replication factor of the above topic is set to 4, then Kafka will create four identical replicas of each partition and place them in the cluster to make them available for all its operations. Replicas are never used to read or write data. They are used to prevent data loss.

Messaging System

A messaging system is a system that is used for transferring data from one application to another so that the applications can focus on data and not on how to share it. Kafka is a distributed publish-subscribe messaging system. In a publish-subscribe system, messages are persisted in a topic. Message producers are called publishers and message consumers are called subscribers. Consumers can subscribe to one or more topic and consume all the messages in that topic.

Two types of messaging patterns are available − one is point to point and the other is publish-subscribe (pub-sub) messaging system. Most of the messaging patterns follow pub-sub.

  • Point to Point Messaging System – In a point-to-point system, messages are persisted in a queue. One or more consumers can consume the messages in the queue, but a particular message can be consumed by a maximum of one consumer only. Once a consumer reads a message in the queue, it disappears from that queue. The typical example of this system is an Order Processing System, where each order will be processed by one Order Processor, but Multiple Order Processors can work as well at the same time.
  • Publish-Subscribe Messaging System – In the publish-subscribe system, messages are persisted in a topic. Unlike point-to-point system, consumers can subscribe to one or more topic and consume all the messages in that topic. In the Publish-Subscribe system, message producers are called publishers and message consumers are called subscribers. A real-life example is Dish TV, which publishes different channels like sports, movies, music, etc., and anyone can subscribe to their own set of channels and get them whenever their subscribed channels are available


Brokers are simple systems responsible for maintaining published data. Kafka brokers are stateless, so they use ZooKeeper for maintaining their cluster state. Each broker may have zero or more partitions per topic. For example, if there are 10 partitions on a topic and 10 brokers, then each broker will have one partition. But if there are 10 partitions and 15 brokers, then the starting 10 brokers will have one partition each and the remaining five won’t have any partition for that particular topic. However, if partitions are 15 but brokers are 10, then brokers would be sharing one or more partitions among them, leading to unequal load distribution among the brokers. Try to avoid this scenario.


When Kafka has more than one broker, it is called a Kafka cluster. A Kafka cluster can be expanded without downtime. These clusters are used to manage the persistence and replication of message data.
You can deploy Kafka as a multi-tenant solution. Multi-tenancy is enabled by configuring which topics can produce or consume data. There is also operations support for quotas. Administrators can define and enforce quotas on requests to control the broker resources that are used by clients. For more information, see the security documentation.


ZooKeeper is used for managing and coordinating Kafka brokers. ZooKeeper is mainly used to notify producers and consumers about the presence of any new broker in the Kafka system or about the failure of any broker in the Kafka system. ZooKeeper notifies the producer and consumer about the presence or failure of a broker based on which producer and consumer makes a decision and starts coordinating their tasks with some other broker.

2. Architecture

Kafka has four core APIs:

  • The Producer API allows an application to publish a stream of records to one or more Kafka topics.
  • The Consumer API allows an application to subscribe to one or more topics and process the stream of records produced to them.
  • The Streams API allows an application to act as a stream processor, consuming an input stream from one or more topics and producing an output stream to one or more output topics, effectively transforming the input streams to output streams.
  • The Connector API allows building and running reusable producers or consumers that connect Kafka topics to existing applications or data systems. For example, a connector to a relational database might capture every change to a table.

Apache describes Kafka as a distributed streaming platform that lets us:

  • Publish and subscribe to streams of records.
  • Store streams of records in a fault-tolerant way.
  • Process streams of records as they occur. states that:

  • Kafka runs as a cluster on one or more servers.
  • The Kafka cluster stores a stream of records in categories called topics.
  • Each record consists of a key, a value, and a timestamp.

3. History

Kafka was developed around 2010 at LinkedIn by a team that included Jay Kreps, Jun Rao, and Neha Narkhede. The problem they originally set out to solve was low-latency ingestion of large amounts of event data from the LinkedIn website and infrastructure into a lambda architecture that harnessed Hadoop and real-time event processing systems. The key was the “real-time” processing. At the time, there weren’t any solutions for this type of ingress for real-time applications.

There were good solutions for ingesting data into offline batch systems, but they exposed implementation details to downstream users and used a push model that could easily overwhelm a consumer. Also, they were not designed for the real-time use case.

Kafka was developed to be the ingestion backbone for this type of use case. Back in 2011, Kafka was ingesting more than 1 billion events a day. Recently, LinkedIn has reported ingestion rates of 1 trillion messages a day.

Why Kafka?

In Big Data, an enormous volume of data is used. But how are we going to collect this large volume of data and analyze that data? To overcome this, we need a messaging system. That is why we need Kafka. The functionalities that it provides are well-suited for our requirements, and thus we use Kafka for:

  • Building real-time streaming data pipelines that can get data between systems and applications.
  • Building real-time streaming applications to react to the stream of data.

Kafka can work with Flume/Flafka, Spark Streaming, Storm, HBase, Flink, and Spark for real-time ingesting, analysis and processing of streaming data. Kafka is a data stream used to feed Hadoop Big Data lakes. Kafka brokers support massive message streams for low-latency follow-up analysis in Hadoop or Spark. Also, Kafka Streaming (a subproject) can be used for real-time analytics.

Why is it so popular? published an article in February 2016 documenting some interesting stats around the “rise and rise” of a powerful asynchronous messaging technology called Apache Kafka.

Kafka has operational simplicity. Kafka is to set up and use, and it is easy to figure out how Kafka works. However, the main reason Kafka is very popular is its excellent performance. It is stable, provides reliable durability, has a flexible publish-subscribe/queue that scales well with N-number of consumer groups, has robust replication, provides producers with tunable consistency guarantees, and it provides preserved ordering at the shard level (i.e. Kafka topic partition). In addition, Kafka works well with systems that have data streams to process and enables those systems to aggregate, transform, and load into other stores. But none of those characteristics would matter if Kafka was slow. The most important reason Kafka is popular is Kafka’s exceptional performance.

Who Uses Kafka?

A lot of large companies who handle a lot of data use Kafka. LinkedIn, where it originated, uses it to track activity data and operational metrics. Twitter uses it as part of Storm to provide a stream processing infrastructure. Square uses Kafka as a bus to move all system events to various Square data centers (logs, custom events, metrics, and so on), outputs to Splunk, for Graphite (dashboards), and to implement Esper-like/CEP alerting systems. It’s also used by other companies like Spotify, Uber, Tumbler, Goldman Sachs, PayPal, Box, Cisco, CloudFlare, and Netflix.

Why Is Kafka So Fast?

Kafka relies heavily on the OS kernel to move data around quickly. It relies on the principals of zero copy. Kafka enables you to batch data records into chunks. These batches of data can be seen end-to-end from producer to file system (Kafka topic log) to the consumer. Batching allows for more efficient data compression and reduces I/O latency. Kafka writes to the immutable commit log to the disk sequential, thus avoiding random disk access and slow disk seeking. Kafka provides horizontal scale through sharding. It shards a topic log into hundreds (potentially thousands) of partitions to thousands of servers. This sharding allows Kafka to handle massive load.

Benefits of Kafka

Four main benefits of Kafka are:

  • Reliability. Kafka is distributed, partitioned, replicated, and fault tolerant. Kafka replicates data and is able to support multiple subscribers. Additionally, it automatically balances consumers in the event of failure.
  • Scalability. Kafka is a distributed system that scales quickly and easily without incurring any downtime.
  • Durability. Kafka uses a distributed commit log, which means messages persists on disk as fast as possible providing intra-cluster replication, hence it is durable.
  • Performance. Kafka has high throughput for both publishing and subscribing messages. It maintains stable performance even when dealing with many terabytes of stored messages.

Use Cases

Kafka can be used in many Use Cases. Some of them are listed below

  • Metrics − Kafka is often used for operational monitoring data. This involves aggregating statistics from distributed applications to produce centralized feeds of operational data.
  • Log Aggregation Solution − Kafka can be used across an organization to collect logs from multiple services and make them available in a standard format to multiple con-sumers.
  • Stream Processing − Popular frameworks such as Storm and Spark Streaming read data from a topic, processes it, and write processed data to a new topic where it becomes available for users and applications. Kafka’s strong durability is also very useful in the context of stream processing.

4. Courses

5. Book

Kafka: The Definitive Guide is the best option to start.


6. Influencers List




7. Link


Apache Kafka

Thorough Introduction to Apache Kafka

A good Kafka explanation

What is Kafka

Kafka Architecture and Its Fundamental Concepts

Apache Kafka Tutorial — Kafka For Beginners

What to consider for painless Apache Kafka integration

What is Data Lake?

How’s it going there?

if you’re not familiar with Big Data, I suggest you have a look on my post “What is Big Data?
” before.
This post is a collection of links, videos, tutorials, blogs and books that I found mixed with my opinion.

Table of contents

1. What is Data Lake?
2. History
3. courses
4. Books
5. Influencers List
6. Link

1.What is Data Lake?

Like Big Data is something no stratforward to explain and there’s no unique answer to that. Even though there is no single definition for Data Lake that is universally accepted, there are some common concepts and I’ll try to cover in this post.

I like the simple definition:
Data lake is a place to store your structured and unstructured data, as well as a method for organizing large volumes of highly diverse data from diverse sources.

I Googled about and I found a different answer.
A data lake is a massive, easily accessible, centralized repository of large volumes of structured and unstructured data. The data lake architecture is a store-everything approach to big data. Data are not classified when they are stored in the repository, as the value of the data is not clear at the outset. As a result, data preparation is eliminated. A data lake is thus less structured compared to a conventional data warehouse. When the data are accessed, only then are they classified, organized or analyzed.

Other answer.
A data lake is a collection of storage instances of various data assets additional to the originating data sources. These assets are stored in a near-exact, or even exact, copy of the source format. The purpose of a data lake is to present an unrefined view of data to only the most highly skilled analysts, to help them explore their data refinement and analysis techniques independent of any of the system-of-record compromises that may exist in a traditional analytic data store (such as a data mart or data warehouse).

According to Nick Huedecker at Gartner,
Data lakes are marketed as enterprise-wide data management platforms for analyzing disparate sources of data in its native format. The idea is simple: instead of placing data in a purpose-built data store, you move it into a data lake in its original format. This eliminates the upfront costs of data ingestion, like transformation. Once data is placed into the lake, it’s available for analysis by everyone in the organization.

Forbes try to explain Data Lake making a comparison with Data Warehouse:

Martin Fowler.
The idea is to have a single store for all of the raw data that anyone in an organisation might need to analyse. Commonly people use Hadoop to work on the data in the lake, but the concept is broader than just Hadoop.

Data Lakes by Oracle

Data lakes are becoming increasingly important as people, especially in business and technology, want to perform broad data exploration and discovery. Bringing data together into a single place or most of it in a single place can be useful for that.
Most data lake implementations are probably based on the Hadoop ecosystem, which is a set of tools that makes it easy to use MapReduce or other computation models.
All data lakes have some distributed file systems. Data should be persisted in raw format because it’s not possible to structure them on ingestion. To achieve this, ingested data should be left in raw form; later they can be structured with transformation processes. As you can see, there is a need for a dedicated layer which allows unstructured data to persist efficiently. In Hadoop, HDFS fulfills this role.

To build ingestion and transformation processes, we need to use some computation system that is fault-tolerant, easily scalable, and efficient at processing large data sets. Nowadays, streaming systems are gaining in popularity. Spark, Storm, Flink… At the beginning of BigData, only MapReduce was available, which was (and still is) used as a bulk-processing framework.
Scalability in a computation system requires resource management. In a data lake, we have huge amounts of data requiring thousands of nodes. Prioritization is achieved by allocating resources and queuing tasks. Some transformations require more resources; some require less. Major tasks get more resources. This resources allocation role in Hadoop is performed by YARN.

What is Streaming?

Streaming is data that is continuously generated by different sources. Such data should be processed incrementally using Stream Processing techniques without having access to all of the data. In addition, it should be considered that concept drift may happen in the data which means that the properties of the stream may change over time.
It is usually used in the context of big data in which it is generated by many different sources at high speed.
Data streaming can also be explained as a technology used to deliver content to devices over the internet, and it allows users to access the content immediately, rather than having to wait for it to be downloaded. Big data is forcing many organizations to focus on storage costs, which brings interest to data lakes and data streams.

In Big Data management, data streaming is the continuous high-speed transfer of large amounts of data from a source system to a target. By efficiently processing and analyzing real-time data streams to glean business insight, data streaming can provide up-to-the-second analytics that enable businesses to quickly react to changing conditions

What can Data Lake do?

This is not an exhaustive list!

  • Ingestion of semi-structured and unstructured data sources (aka big data) such as equipment readings, telemetry data, logs, streaming data, and so forth. A data lake is a great solution for storing IoT (Internet of Things) type of data which has traditionally been more difficult to store, and can support near real-time analysis. Optionally, you can also add structured data (i.e., extracted from a relational data source) to a data lake if your objective is a single repository of all data to be available via the lake.
  • Experimental analysis of data before its value or purpose has been fully defined. Agility is important for every business these days, so a data lake can play an important role in “proof of value” type of situations because of the “ELT” approach discussed above.
  • Advanced analytics support. A data lake is useful for data scientists and analysts to provision and experiment with data.
  • Archival and historical data storage. Sometimes data is used infrequently, but does need to be available for analysis. A data lake strategy can be very valuable to support an active archive strategy.
  • Support for Lambda architecture which includes a speed layer, batch layer, and serving layer.
  • Preparation for data warehousing. Using a data lake as a staging area of a data warehouse is one way to utilize the lake, particularly if you are getting started.
  • Augment a data warehouse. A data lake may contain data that isn’t easily stored in a data warehouse, or isn’t queried frequently. The data lake might be accessed via federated queries which make its separation from the DW transparent to end users via a data virtualization layer.
  • Distributed processing capabilities associated with a logical data warehouse.
  • Storage of all organizational data to support downstream reporting & analysis activities. Some organizations wish to achieve a single storage repository for all types of data. Frequently, the goal is to store as much data as possible to support any type of analysis that might yield valuable findings.
  • Application support. In addition to analysis by people, a data lake can be a data source for a front-end application. The data lake might also act as a publisher for a downstream application (though ingestion of data into the data lake for purposes of analytics remains the most frequently cited use).

Under the umbrella of Data Lake there are many of technologies and concepts. This is not an exhaustive list!

Data ingestion

Data ingestion is the process of flowing data from its origin to one or more data stores, such as a data lake, though this can also include databases and search engines.
Sources can be clickstreams, data center logs, sensors, APIs or even databases. They use various data formats (structured, unstructured, semi-structured, multi-structured), can make data available in a stream or batches, and support various protocols for data movement.

There are two processing frameworks which “ingest” data into data lakes:

  • Batch processing – Millions of blocks of data processed over long periods of time (hours-to-days).
  • Stream processing – Small batches of data processed in real-time. Stream processing is becoming increasingly valuable for businesses that harness real-time analytics.

Data Marts

A Data Mart is an archive of stored, normally structured data, typically used and controlled by a specific community or department. It is normally smaller and more focused than a Data Warehouse and, currently, is often a subdivision of Data Warehouses. Data Marts were the first evolutionary step in the physical reality of Data Warehouses and Data Lakes

Data Silos

Data Silos are part of a Data Warehouse and similar to Data Marts, but much more isolated. Data Silos are insulated management systems that cannot work with other systems. A Data Silo contains fixed data that is controlled by one department and is cut off from other parts of the organization. They tend to form within large organizations due to the different goals and priorities of individual departments. Data Silos also form when departments compete with one another instead of working as a team toward common business goals.

Data Warehouses

Data Warehouses are centralized repositories of information that can be researched for purposes of making better informed decisions. The data comes from a wide range of sources and is often unstructured. Data is accessed through the use of business intelligence tools, SQL clients, and other Analytics applications.

Data swamp

In data lakes there are many things going on and it’s not possible to manage them manually. Without constraints and a thoughtful approach to processes, a data lake will become degenerated very quickly. If ingested data do not contain business information, then we can’t find the right context for them. If everyone generates anonymous data without lineage, then we will have tons of useless data. No one will know what is going on. Who is the author of changes? Where did the data come from? Everything starts to look like a data swamp.

Data Mining

Data mining is defined as “knowledge discovery in databases,” and is how data scientists uncover previously unseen patterns and truths through various models. data mining is about sifting through large sets of data to uncover patterns, trends, and other truths that may not have been previously visible.

Data Cleansing

The goal of data cleansing is to improve data quality and utility by catching and correcting errors before it is transferred to a target database or data warehouse. Manual data cleansing may or may not be realistic, depending on the amount of data and number of data sources your company has. Regardless of the methodology, data cleansing presents a handful of challenges, such as correcting mismatches, ensuring that columns are in the same order, and checking that data (such as date or currency) is in the same format.

Data Quality

People try to describe data quality using terms like complete, accurate, accessible, and de-duped. And while each of these words describes a specific element of data quality, the larger concept of data quality is really about whether or not that data fulfills the purpose or purposes you want to use it for.


ETL stands for “Extract, Transform, Load”, and is the common paradigm by which data from multiple systems is combined to a single database, data store, or warehouse for legacy storage or analytics.


ELT is a process that involves extracting the data, loading it to the target warehouse, and then transforming it after it is loaded. In this case, the work of transforming the data is completed by the target database.

Data extraction

Data extraction is a process that involves retrieval of data from various sources. Frequently, companies extract data in order to process it further, migrate the data to a data repository (such as a data warehouse or a data lake) or to further analyze it. It’s common to transform the data as a part of this process.

Data loading

Data loading refers to the “load” component of ETL. After data is retrieved and combined from multiple sources (extracted), cleaned and formatted (transformed), it is then loaded into a storage system, such as a cloud data warehouse.

Data Transformation

Data transformation is the process of converting data from one format or structure into another format or structure. Data transformation is critical to activities such as data integration and data management. Data transformation can include a range of activities: you might convert data types, cleanse data by removing nulls or duplicate data, enrich the data, or perform aggregations, depending on the needs of your project.

Data Migration

Data migration is simply the process of moving data from a source system to a target system. Companies have many different reasons for migrating data. You may want to migrate data when you acquire another company and you need to integrate that company’s data. Or, you may want to integrate data from different departments within your company so the data is available across your entire business. You may want to move your data from an on-premise platform to a cloud platform. Or, perhaps you’re moving from an outdated data storage system to a new database or data storage system. The concept of data migration is simple, but it can sometimes be a complex process.

Lambda architecture

Lambda architecture is a data-processing architecture designed to handle massive quantities of data by taking advantage of both batch and stream-processing methods. This approach to architecture attempts to balance latency, throughput, and fault-tolerance by using batch processing to provide comprehensive and accurate views of batch data, while simultaneously using real-time stream processing to provide views of online data. The two view outputs may be joined before presentation. The rise of lambda architecture is correlated with the growth of big data, real-time analytics, and the drive to mitigate the latencies of map-reduce.

Kappa Architecture

Kappa Architecture is a software architecture pattern. Rather than using a relational DB like SQL or a key-value store like Cassandra, the canonical data store in a Kappa Architecture system is an append-only immutable log. From the log, data is streamed through a computational system and fed into auxiliary stores for serving.
Kappa Architecture is a simplification of Lambda Architecture. A Kappa Architecture system is like a Lambda Architecture system with the batch processing system removed. To replace batch processing, data is simply fed through the streaming system quickly.

SIEM Security information and event management

The underlying principles of every SIEM system is to aggregate relevant data from multiple sources, identify deviations from the norm and take appropriate action. For example, when a potential issue is detected, a SIEM might log additional information, generate an alert and instruct other security controls to stop an activity’s progress.

Security information and event management (SIEM) is an approach to security management that combines SIM (security information management) and SEM (security event management) functions into one security management system. The acronym SIEM is pronounced “sim” with a silent e.

software collects and aggregates log data generated throughout the organization’s technology infrastructure, from host systems and applications to network and security devices such as firewalls and antivirus filters.The software then identifies and categorizes incidents and events, as well as analyzes them.

SNA social network analysis

Social network analysis is the process of investigating social structures through the use of networks and graph theory. It characterizes networked structures in terms of nodes (individual actors, people, or things within the network) and the ties, edges, or links (relationships or interactions) that connect them.

Data lake vs data warehouse

Data warehouses rely on structure and clean data, whereas data lakes allow data to be in its most natural form. This is because advanced analytic tools and mining software intake raw data and transform it into useful insight.

Both data lakes and data warehouses are repositories for data. That’s about the only similarity between the two. Now, let’s touch on some of the key differences:

  • Data lakes are designed to support all types of data, whereas data warehouses make use of highly structured data – in most cases.
  • Data lakes store all data that may or may not be analyzed at some point in the future. This principle doesn’t apply to data warehouses since irrelevant data is typically eliminated due to limited storage.
  • The scale between data lakes and data warehouses is drastically different due to our previous points. Supporting all types of data and storing that data (even if it’s not immediately useful) means data lakes need to be highly scalable.
  • Thanks to metadata (data about data), users working with a data lake can gain basic insight about the data quickly. In data warehouses, it often requires a member of the development team to access the data – which could create a bottleneck.
  • Lastly, the intense data management required for data warehouses means they’re typically more expensive to maintain compared to data lakes.

I’m believer that modern data warehousing is still very important. Therefore, a data lake itself, doesn’t entirely replace the need for a data warehouse (or data marts) which contain cleansed data in a user-friendly format. The data warehouse doesn’t absolutely have to be in a relational database anymore, but it does need a semantic layer which is easy to work with that most business users can access for the most common reporting needs.

There’s always tradeoffs between performing analytics on a data lake versus from a cleansed data warehouse: Query performance, data load performance, scalability, reusability, data quality, and so forth. Therefore, I believe that a data lake and a data warehouse are complementary technologies that can offer balance. For a fast analysis by a highly qualified analyst or data scientist, especially exploratory analysis, the data lake is appealing. For delivering cleansed, user-friendly data to a large number of users across the organization, the data warehouse still rules.

2. History

Data Lakes allow Data Scientists to mine and analyze large amounts of Big Data. Big Data, which was used for years without an official name, was labeled by Roger Magoulas in 2005. He was describing a large amount of data that seemed impossible to manage or research using the traditional SQL tools available at the time. Hadoop (2008) provided the search engine needed for locating and processing unstructured data on a massive scale, opening the door for Big Data research.

In October of 2010, James Dixon, founder and former CTO of Pentaho, came up with the term “Data Lake”

Data Lake came from the idea that the data drop in one place, and this place becomes a lake.
Visiting a large lake is always a very pleasant feeling. The water in the lake is in its purest form and there are different activities different people perform on the Lake. Some are people are fishing, some people are enjoying a boat ride, this lake also supplies drinking water to people. In short, the same lake is used for multiple purposes. Data Lake Architecture, like the water in the lake, data in a data lake is in the purest possible form. Like the lake, it caters to need to different people, those who want to fish or those who want to take a boat ride or those who want to get drinking water from it, a data lake architecture caters to multiple personas.

This is not a official history about how the name came from. If you know please leave one comment.

3. Courses

4. Books

The Enterprise Big Data Lake

Data Lake for Enterprises

5. Influencers List

6. Link

Data lakes and the promise of unsiloed data

Every software engineer should know about real time datas

Questioning the lambda architecture

Whats a data lake

Data Lake architecture

What’s the Difference Between a Data Lake, Data Warehouse and Database?

What is data streaming

What is Apache Spark?

How heya?

if you’re not familiar with Big Data, I suggest you have a look on my post “What is Big Data?” before.
This post is a collection of links, videos, tutorials, blogs and books that I found mixed with my opinion.

Table of contents

1. What is Apache Spark?
2. Architecture
3. History
4. Courses
5. Books
6. Influencers List
7. Link

1. What is Apache Spark?

Apache Spark is an open source parallel processing framework for storing and processing Big Data across clustered computers. Spark can be used to perform computations much faster than what Hadoop can rather Hadoop and Spark can be used together efficiently. Spark is written in Scala, which is considered the primary language for interacting with the Spark Core engine, but it doesn’t require developers to know Scala, which executes inside a Java Virtual Machine (JVM). APIs for Java, Python, R, and Scala ensure Spark is within reach of a wide audience of developers, and they have embraced the software.

Apache Spark is most actively developed open source project in big data and probably the most widely used as well. Spark is a general purpose Execution Engine, which can perform its cluster management on top of Big Data very quickly and efficiently. It is rapidly increasing its features and capabilities like libraries to perform different types of Analytics.

Apache Spark is a fast, in-memory data processing engine with elegant and expressive development APIs to allow data workers to efficiently execute streaming, machine learning or SQL workloads that require fast iterative access to datasets. With Spark running on Apache Hadoop YARN, developers everywhere can now create applications to exploit Spark’s power, derive insights, and enrich their data science workloads within a single, shared dataset in Hadoop.
Apache Spark is a lightning-fast cluster computing technology, designed for fast computation. It is based on Hadoop MapReduce and it extends the MapReduce model to efficiently use it for more types of computations, which includes interactive queries and stream processing. The main feature of Spark is its in-memory cluster computing that increases the processing speed of an application

The Hadoop YARN-based architecture provides the foundation that enables Spark and other applications to share a common cluster and dataset while ensuring consistent levels of service and response. Spark is now one of many data access engines that work with YARN in HDP.
Spark is designed for data science and its abstraction makes data science easier.

Spark also includes MLlib, a library that provides a growing set of machine algorithms for common data science techniques: Classification, Regression, Collaborative Filtering, Clustering and Dimensionality Reduction.

The Driver and the Executer

Spark uses a master-slave architecture. A driver coordinates many distributed workers in order to execute tasks in a distributed manner while a resource manager deals with the resource allocation to get the tasks done.


Think of it as the “Orchestrator”. The driver is where the main method runs. It converts the program into tasks and then schedules the tasks to the executors. The driver has at its disposal 3 different ways of communicating with the executors; Broadcast, Take and DAG. It controls the execution of a Spark application and maintains all of the states of the Spark cluster, which includes the state and tasks of the executors. The driver must interface with the cluster manager in order to get physical resources and launch executors. To put this in simple terms, this process is just a process on a physical machine that is responsible for maintaining the state of the application running on the cluster.

  • Broadcast Action: The driver transmits the necessary data to each executor. This action is optimal for data sets under a million records, +- 1gb of data. This action can become a very expensive task.
  • Take Action: Driver takes data from all Executors. This action can be a very expensive and dangerous action as the driver might run out of memory and the network could become overwhelmed.
  • DAG Action: (Direct Acyclic Graph) This is the by far least expensive action out of the three. It transmits control flow logic from the driver to the executors.

Executer –  “Workers”

Executers execute the delegated tasks from the driver within a JVM instance. Executors are launched at the beginning of a Spark application and normally run for the whole life span of an application. This method allows for data to persist in memory while different tasks are loaded in and out of the execute throughout the application’s lifespan.
The JVM worker environments in Hadoop MapReduce in stark contrast powers down and powers up for each task. The consequence of this is that Hadoop must perform reads and writes on disk at the start and end of every task.

Cluster manager

This is responsible for allocating resources across the spark application. The Spark context is capable of connecting to several types of cluster managers like Mesos, Yarn or Kubernetes apart from the Spark’s standalone cluster manager.
Cluster Manager is responsible for maintaining a cluster of machines that will run your Spark Application. Cluster managers have their own ‘driver’ and ‘worker’ abstractions, but the difference is that these are tied to physical machines rather than processes.

Spark Context

It holds a connection with Spark cluster manager. All Spark applications run as independent set of processes, coordinated by a SparkContext in a program.

Spark has three data representations viz RDD, Dataframe, Dataset. For each data representation, Spark has a different API. Dataframe is much faster than RDD because it has metadata (some information about data) associated with it, which allows Spark to optimize query plan.
A nice place to understand more about RDD, Dataframe, Dataset is this article: “A Tale of Three Apache Spark APIs: RDDs vs DataFrames and Datasets”


A Resilient Distributed Dataset (RDD), is the primary data abstraction in Apache Spark and the core of Spark. Represents an immutable, partitioned collection of elements that can be operated on in parallel. This class contains the basic operations available on all RDDs, such as map, filter, and persist. Is the most basic data abstraction in Spark. It is a fault-tolerant collection of elements that can be operated on in parallel.RDD is an immutable distributed collection of elements of your data, partitioned across nodes in your cluster that can be operated in parallel with a low-level API that offers transformations and actions.

We can apply 2 types of operations on RDDs

Transformation: Transformation refers to the operation applied on a RDD to create new RDD.
Action: Actions refer to an operation which also apply on RDD that perform computation and send the result back to driver.

Example: Map (Transformation) performs operation on each element of RDD and returns a new RDD. But, in case of Reduce (Action), it reduces / aggregates the output of a map by applying some functions (Reduce by key).

RDDs use Shared Variable

The parallel operations in Apache Spark use shared variable. It means that whenever a task is sent by a driver to executors program in a cluster, a copy of shared variable is sent to each node in a cluster, so that they can use this variable while performing task. Accumulator and Broadcast are the two types of shared variables supported by Apache Spark.
Broadcast: We can use the Broadcast variable to save the copy of data across all node.
Accumulator: In Accumulator variables are used for aggregating the information.

How to Create RDD in Apache Spark

Existing storage: When we want to create a RDD though existing storage in driver program (which we would like to be parallelized). For example, converting a list to RDD, which is already created in a driver program.

External sources: When we want to create a RDD though external sources such as a shared file system, HDFS, HBase, or any data source offering a Hadoop Input Format.

The features of RDDs :

  • Resilient, i.e. fault-tolerant with the help of RDD lineage graph and so able to recompute missing or damaged partitions due to node failures.
  • Distributed with data residing on multiple nodes in a cluster.
  • Dataset is a collection of partitioned data with primitive values or values of values, e.g. tuples or other objects (that represent records of the data you work with).

The key reasons RDDs are an abstraction that works better for distributed data processing, is because they don’t feature some of the issues that MapReduce, the older paradigm for data processing (which Spark is replacing increasingly). Chiefly, these are:

1. Replication: Replication of data on different parts of a cluster, is a feature of HDFS that enables data to be stored in a fault-tolerant manner. Spark’s RDDs address fault tolerance by using a lineage graph. The different name (resilient, as opposed to replicated) indicates this difference of implementation in the core functionality of Spark

2. Serialization: Serialization in MapReduce bogs it down, speed wise, in operations like shuffling and sorting.


It is a distributed collection of data organized into named columns. It is conceptually equivalent to a table in a relational database or a data frame but with lot more stuff under the hood. DataFrame allows developers to impose a structure onto a distributed collection of data, allowing higher-level abstractions. From 2.0, the Data frames API was merged with Dataset. Now a DataFrame is a Dataset organized into named columns.
Data frames can be created from RDDs.
Dataframes also tries to solve a lot of performance issues that spark had with non-jvm languages like python and R . Historically, using RDD’s in python was much slower than in Scala. With Dataframes, code written all the languages perform the same with some exceptions.


It is a collection of strongly-typed domain-specific objects that can be transformed in parallel using functional or relational operations. A logical plan is created and updated for each transformation and a final logical plan is converted to a physical plan when an action is invoked. Spark’s catalyst
optimizer optimizes the logical plan and generates a physical plan for efficient execution in a parallel and distributed manner. Further, there are Encoders generates optimized, lower memory footprint binary structure. Encoders know the schema of the records. This is how they offer significantly faster
serialization and deserialization (comparing to the default Java or Kryo serializers).

The main advantage of Datasets is Type safety. When using Datasets we are assured that both the syntax errors and Analysis errors are caught during compile time. In contrast with Dataframes, where a syntax error can be caught during compile time but an Analysis error such as referring to a nonexisting column name would be caught only once you run it. The run times can be quite expensive and also as a developer it would be nice to have compiler and IDE’s to do these jobs for you.

2. Architecture

Spark Core

Spark Core is the base engine for large-scale parallel and distributed data processing. Further, additional libraries which are built on the top of the core allows diverse workloads for streaming, SQL, and machine learning. It is responsible for memory management and fault recovery, scheduling, distributing and monitoring jobs on a cluster & interacting with storage systems.


BlindDB or Blind Database is also known as an Approximate SQL database. If there is a huge amount of data barraging and you are not really interested in accuracy, or in exact results, but just want to have a rough or an approximate picture, BlindDB gets you the same. Firing a query, doing some sort of sampling, and giving out some output is called Approximate SQL. Isn’t it a new and interesting concept? Many a time, when you do not require accurate results, sampling would certainly do.

Spark SQL

Spark SQL is a new module in Spark which integrates relational processing with Spark’s functional programming API. It supports querying data either via SQL or via the Hive Query Language. For those of you familiar with RDBMS, Spark SQL will be an easy transition from your earlier tools where you can extend the boundaries of traditional relational data processing.

Spark Streaming

Spark Streaming is one of those unique features, which have empowered Spark to potentially take the role of Apache Storm. Spark Streaming mainly enables you to create analytical and interactive applications for live streaming data. You can do the streaming of the data and then, Spark can run its operations from the streamed data itself.

Structured Streaming

Structured Streaming (added in Spark 2.x) is to Spark Streaming what Spark SQL was to the Spark Core APIs: A higher-level API and easier abstraction for writing applications. In the case of Structure Streaming, the higher-level API essentially allows developers to create infinite streaming dataframes and datasets. It also solves some very real pain points that users have struggled with in the earlier framework, especially concerning dealing with event-time aggregations and late delivery of messages. All queries on structured streams go through the Catalyst query optimizer, and can even be run in an interactive manner, allowing users to perform SQL queries against live streaming data.
Structured Streaming is still a rather new part of Apache Spark, having been marked as production-ready in the Spark 2.2 release. However, Structured Streaming is the future of streaming applications with the platform, so if you’re building a new streaming application, you should use Structured Streaming. The legacy Spark Streaming APIs will continue to be supported, but the project recommends porting over to Structured Streaming, as the new method makes writing and maintaining streaming code a lot more bearable.


MLLib is a machine learning library like Mahout. It is built on top of Spark, and has the provision to support many machine learning algorithms. But the point difference with Mahout is that it runs almost 100 times faster than MapReduce. It is not yet as enriched as Mahout, but it is coming up pretty well, even though it is still in the initial stage of growth.


For graphs and graphical computations, Spark has its own Graph Computation Engine, called GraphX. It is similar to other widely used graph processing tools or databases, like Neo4j, Girafe, and many other distributed graph databases.

3. History

Apache Spark is about to turn 10 years old.
Spark started in 2009 as a research project in the UC Berkeley RAD Lab, later to become the AMPLab. The researchers in the lab had previously been working on Hadoop MapReduce, and observed that MapReduce was inefficient for iterative and interactive computing jobs. Thus, from the beginning, Spark was designed to be fast for interactive queries and iterative algorithms, bringing in ideas like support for in-memory storage and efficient fault recovery.

Soon after its creation it was already 10–20× faster than MapReduce for certain jobs.
Some of Spark’s first users were other groups inside UC Berkeley, including machine learning researchers such as the Mobile Millennium project, which used Spark to monitor and predict traffic congestion in the San Francisco Bay Area.

Spark was first open sourced in March 2010, and was transferred to the Apache Software Foundation in June 2013, where it is now a top-level project. It is an open source project that has been built and is maintained by a thriving and diverse community of developers. In addition to UC Berkeley, major contributors to Spark include Databricks, Yahoo!, and Intel.

Internet powerhouses such as Netflix, Yahoo, and eBay have deployed Spark at massive scale, collectively processing multiple petabytes of data on clusters of over 8,000 nodes. It has quickly become the largest open source community in big data, with over 1000 contributors from 250+ organizations.

In-memory computation

The biggest advantage of Apache Spark comes from the fact that it saves and loads the data in and from the RAM rather than from the disk (Hard Drive). If we talk about memory hierarchy, RAM has much higher processing speed than Hard Drive (illustrated in figure below). Since the prices of memory has come down significantly in last few years, in-memory computations have gained a lot of momentum.
Spark uses in-memory computations to speed up 100 times faster than Hadoop framew

Spark VS Hadoop


  • Apache Spark — it’s a lightning-fast cluster computing tool. Spark runs applications up to 100x faster in memory and 10x faster on disk than Hadoop by reducing the number of read-write cycles to disk and storing intermediate data in-memory.
  • Hadoop MapReduce — MapReduce reads and writes from disk, which slows down the processing speed and overall efficiency.

Ease of Use

  • Apache Spark — Spark’s many libraries facilitate the execution of lots of major high-level operators with RDD (Resilient Distributed Dataset).
  • Hadoop — In MapReduce, developers need to hand-code every operation, which can make it more difficult to use for complex projects at scale.

Handling Large Sets of Data

  • Apache Spark — since Spark is optimized for speed and computational efficiency by storing most of the data in memory and not on disk, it can underperform Hadoop MapReduce when the size of the data becomes so large that insufficient RAM becomes an issue.
  • Hadoop — Hadoop MapReduce allows parallel processing of huge amounts of data. It breaks a large chunk into smaller ones to be processed separately on different data nodes. In case the resulting dataset is larger than available RAM, Hadoop MapReduce may outperform Spark. It’s a good solution if the speed of processing is not critical and tasks can be left running overnight to generate results in the morning.

Hadoop vs Spark
How do Hadoop and Spark Stack Up?

4. Courses

5. Books

Spark: The Definitive Guide is the best option to start.

Learning Spark: Lightning-Fast Big Data Analysis

Advanced Analytics with Spark: Patterns for Learning from Data at Scale

High Performance Spark: Best Practices for Scaling and Optimizing Apache Spark

6. Influencers List


7. Link

Practical apache spark 10 minutes

Apache Spark Architecture

Structured streaming

Churn prediction with Spark

Introduction to Spark Graphframe

Spark Tutorial

What is Hadoop?

Alright, Boyo?

if you’re not familiar with Big Data, I suggest you have a look on my post “What is Big Data?
” before.
This post is a collection of links, videos, tutorials, blogs and books that I found mixed with my opinion.

Table of contents

1. What is Hadoop?
2. Architecture
3. History
4. Courses
5. Books
6. Influence’s List
7. Podcasts
8. Newsletters
9. Links

1. What is Hadoop?

Hadoop is a framework that allows you to first store Big Data in a distributed environment, so that, you can process it parallelly.
Apache Hadoop is an open source software framework for storage and large scale processing of data-sets on clusters of commodity hardware. Hadoop is an Apache top-level project being built and used by a global community of contributors and users. It is licensed under the Apache License 2.0.

According to the definition of 3V’s of big data, Apache Hadoop came to solve these problems.

The first problem is storing Big data.

HDFS provides a distributed way to store Big data. Your data is stored in blocks across the DataNodes and you can specify the size of blocks. Basically, if you have 512MB of data and you have configured HDFS such that, it will create 128 MB of data blocks. So HDFS will divide data into 4 blocks as 512/128=4 and store it across different DataNodes, it will also replicate the data blocks on different DataNodes. Now, as we are using commodity hardware, hence storing is not a challenge.
It also solves the scaling problem. It focuses on horizontal scaling instead of vertical scaling. You can always add some extra data nodes to HDFS cluster as and when required, instead of scaling up the resources of your DataNodes.

The second problem was storing a variety of data.

With HDFS you can store all kinds of data whether it is structured, semi-structured or unstructured. Since in HDFS, there is no pre-dumping schema validation. And it also follows write once and read many models. Due to this, you can just write the data once and you can read it multiple times for finding insights.

The third challenge was accessing & processing the data faster.

This is one of the major challenges with Big Data. In order to solve it, we move processing to data and not data to processing. What does it mean? Instead of moving data to the master node and then processing it. In MapReduce, the processing logic is sent to the various slave nodes & then data is processed parallely across different slave nodes. Then the processed results are sent to the master node where the results are merged and the response is sent back to the client.
In YARN architecture, we have ResourceManager and NodeManager. ResourceManager might or might not be configured on the same machine as NameNode. But, NodeManagers should be configured on the same machine where DataNodes are present.

2. Architecture

We can divide Hadoop in some modules;

A. Hadoop Common: contains libraries and utilities needed by other Hadoop modules
B. Hadoop Distributed File System (HDFS): a distributed file-system that stores data on the commodity machines, providing very high aggregate bandwidth across the cluster
C. Hadoop MapReduce: a programming model for large scale data processing
D. Hadoop YARN: a resource-management platform responsible for managing compute resources in clusters and using them for scheduling of users’ applications

A. Hadoop Common
Refers to the collection of common utilities and libraries that support other Hadoop modules. It is an essential part or module of the Apache Hadoop Framework, along with the Hadoop Distributed File System (HDFS), Hadoop YARN and Hadoop MapReduce.

B. How the Hadoop Distributed File System (HDFS) works
Hadoop has a file system that is much like the one on your desktop computer, but it allows us to distribute files across many machines. HDFS organizes information into a consistent set of file blocks and storage blocks for each node. In the Apache distribution, the file blocks are 64MB and the storage blocks are 512 KB. Most of the nodes are data nodes, and there are also copies of the data. Name nodes exist to keep track of where all the file blocks reside.

Each node in a Hadoop instance typically has a single namenode, and a cluster of datanodes form the HDFS cluster. The situation is typical because each node does not require a datanode to be present. Each datanode serves up blocks of data over the network using a block protocol specific to HDFS. The file system uses the TCP/IP layer for communication. Clients use Remote procedure call (RPC) to communicate with each other. With the default replication value, 3, data is stored on three nodes: two on the same rack, and one on a different rack. Data nodes can talk to each other to rebalance data, to move copies around, and to keep the replication of data high.

The HDFS file system includes a so-called secondary namenode, which misleads some people into thinking that when the primary namenode goes offline, the secondary namenode takes over. In fact, the secondary namenode regularly connects with the primary namenode and builds snapshots of the primary namenode’s directory information, which the system then saves to local or remote directories.
File access can be achieved through the native Java API, the Thrift API, to generate a client in the language of the users’ choosing ( Java, Python, Scala, …), the command-line interface, or browsed through the HDFS-UI web app over HTTP.

Maintains and Manages DataNodes.
Records Metadata i.e. information about data blocks e.g. location of blocks stored, the size of the files, permissions, hierarchy, etc.
Receives status and block report from all the DataNodes.

Slave daemons. It sends signals to NameNode.
Stores actual It stores in data blocks.
Serves read and write requests from the clients.

C. How MapReduce works

Map Reduce is a really powerful programming model that was built by some smart guys at Google. It helps to process really large sets of data on a cluster using a parallel distributed algorithm.

As the name suggests, there are two steps in the MapReduce process—map and reduce. Let’s say you start with a file containing all the blog entries about big data in the past 24 hours and want to count how many times the words Hadoop, Big Data, and Greenplum are mentioned. First, the file gets split up on HDFS. Then, all participating nodes go through the same map computation for their local dataset—they count the number of times these words show up. When the map step is complete, each node outputs a list of key-value pairs.
Once mapping is complete, the output is sent to other nodes as input for the reduce step. Before reduce runs, the key-value pairs are typically sorted and shuffled. The reduce phase then sums the lists into single entries for each word.

Above the file systems comes the MapReduce engine, which consists of one JobTracker, to which client applications submit MapReduce jobs. The JobTracker pushes work out to available TaskTracker nodes in the cluster, striving to keep the work as close to the data as possible.

With a rack-aware file system, the JobTracker knows which node contains the data, and which other machines are nearby. If the work cannot be hosted on the actual node where the data resides, priority is given to nodes in the same rack. This reduces network traffic on the main backbone network.

If a TaskTracker fails or times out, that part of the job is rescheduled. The TaskTracker on each node spawns off a separate Java Virtual Machine process to prevent the TaskTracker itself from failing if the running job crashes the JVM. A heartbeat is sent from the TaskTracker to the JobTracker every few minutes to check its status. The Job Tracker and TaskTracker status and information is exposed by Jetty and can be viewed from a web browser.

Some of the terminologies in the MapReduce process are:

MasterNode – Place where JobTracker runs and which accepts job requests from clients
SlaveNode – It is the place where the mapping and reducing programs are run
JobTracker – it is the entity that schedules the jobs and tracks the jobs assigned using Task Tracker
TaskTracker – It is the entity that actually tracks the tasks and provides the report status to the JobTracker
Job – A MapReduce job is the execution of the Mapper & Reducer program across a dataset
Task – the execution of the Mapper & Reducer program on a specific data section
TaskAttempt – A particular task execution attempt on a SlaveNode
Map Function – The map function takes an input and produces a set of intermediate key value pairs.
Reduce Function – The reduce function accepts an Intermediate key and a set of values for that key. It merges together these values to form a smaller set of values. The intermediate values are supplied to user’s reduce function via an iterator.
Thus map reduce converts each task to a group of map reduce functions and each map and reduce task can be performed by different machines. The results can be merged back to produce the required output.

D. How YARN works: Yet Another Resource Negotiator

MapReduce has undergone a complete overhaul in Hadoop 0.23 and we now have, what we call, MapReduce 2.0 (MRv2) or YARN.
Apache Hadoop YARN is a sub-project of Hadoop at the Apache Software Foundation introduced in Hadoop 2.0 that separates the resource management and processing components. YARN was born of a need to enable a broader array of interaction patterns for data stored in HDFS beyond MapReduce. The YARN-based architecture of Hadoop 2.0 provides a more general processing platform that is not constrained to MapReduce.

YARN enhances the power of a Hadoop compute cluster in the following ways:

  • Scalability: The processing power in data centers continues to grow quickly. Because YARN ResourceManager focuses exclusively on scheduling, it can manage those larger clusters much more easily.
  • Compatibility with MapReduce: Existing MapReduce applications and users can run on top of YARN without disruption to their existing processes.
  • Improved cluster utilization: The ResourceManager is a pure scheduler that optimizes cluster utilization according to criteria such as capacity guarantees, fairness, and SLAs. Also, unlike before, there are no named map and reduce slots, which helps to better utilize cluster resources.
  • Support for workloads other than MapReduce: Additional programming models such as graph processing and iterative modeling are now possible for data processing. These added models allow enterprises to realize near real-time processing and increased ROI on their Hadoop investments.
  • Agility: With MapReduce becoming a user-land library, it can evolve independently of the underlying resource manager layer and in a much more agile manner.

The fundamental idea of YARN is to split up the two major responsibilities of the JobTracker/TaskTracker into separate entities:

  • a global ResourceManager
  • a per-application ApplicationMaster
  • a per-node slave NodeManager
  • a per-application container running on a NodeManager

The ResourceManager and the NodeManager form the new, and generic, system for managing applications in a distributed manner. The ResourceManager is the ultimate authority that arbitrates resources among all the applications in the system. The per-application ApplicationMaster is a framework-specific entity and is tasked with negotiating resources from the ResourceManager and working with the NodeManager(s) to execute and monitor the component tasks. The ResourceManager has a scheduler, which is responsible for allocating resources to the various running applications, according to constraints such as queue capacities, user-limits, etc. The scheduler performs its scheduling function based on the resource requirements of the applications. The NodeManager is the per-machine slave, which is responsible for launching the applications’ containers, monitoring their resource usage (CPU, memory, disk, network) and reporting the same to the ResourceManager. Each ApplicationMaster has the responsibility of negotiating appropriate resource containers from the scheduler, tracking their status, and monitoring their progress. From the system perspective, the ApplicationMaster runs as a normal container.


Tez is an extensible framework for building high performance batch and interactive data processing applications, coordinated by YARN in Apache Hadoop. Tez improves the MapReduce paradigm by dramatically improving its speed while maintaining MapReduce’s ability to scale to petabytes of data. Important Hadoop ecosystem projects like Apache Hive and Apache Pig use Apache Tez, as do a growing number of third party data access applications developed for the broader Hadoop ecosystem.

Apache Tez provides a developer API and framework to write native YARN applications that bridge the spectrum of interactive and batch workloads. It allows those data access applications to work with petabytes of data over thousands of nodes. The Apache Tez component library allows developers to create Hadoop applications that integrate natively with Apache Hadoop YARN and perform well within mixed workload clusters.

Since Tez is extensible and embeddable, it provides the fit-to-purpose freedom to express highly optimized data processing applications, giving them an advantage over end-user-facing engines such as MapReduce and Apache Spark. Tez also offers a customizable execution architecture that allows users to express complex computations as dataflow graphs, permitting dynamic performance optimizations based on real information about the data and the resources required to process it.

What is Tez?


MPP stands for Massive Parallel Processing, this is the approach in grid computing when all the separate nodes of your grid are participating in the coordinated computations. MPP DBMSs are the database management systems built on top of this approach. In these systems each query you are staring is split into a set of coordinated processes executed by the nodes of your MPP grid in parallel, splitting the computations the way they are running times faster than in traditional SMP RDBMS systems. One additional advantage that this architecture delivers to you is the scalability, because you can easily scale the grid by adding new nodes into it. To be able to handle huge amounts of data, the data in these solutions is usually split between nodes (sharded) the way that each node processes only its local data. This further speeds up the processing of the data, because using shared storage for this kind of design would be a huge overkill – more complex, more expensive, less scalable, higher network utilization, less parallelism. This is why most of the MPP DBMS solutions are shared-nothing and work on DAS storage or the set of storage shelves shared between small groups of servers.

When to use Hadoop?

Hadoop is used for: (This is not an exhaustive list!)

  • Log processing – Facebook, Yahoo
  • Data Warehouse – Facebook, AOL
  • Video and Image Analysis – New York Times, Eyealike

Till now, we have seen how Hadoop has made Big Data handling possible. But there are some scenarios where Hadoop implementation is not recommended.

When not to use Hadoop?

Following are some of those scenarios : (This is not an exhaustive list!)

  • Low Latency data access : Quick access to small parts of data.
  • Multiple data modification : Hadoop is a better fit only if we are primarily concerned about reading data and not modifying data.
  • Lots of small files : Hadoop is suitable for scenarios, where we have few but large files.
  • After knowing the best suitable use-cases, let us move on and look at a case study where Hadoop has done wonders.

Hate Hadoop? Then You’re Doing It Wrong

3. History

In 2003, Doug Cutting launches project Nutch to handle billions of searches and indexing millions of web pages. Later in Oct 2003 – Google releases papers with GFS (Google File System). In Dec 2004, Google releases papers with MapReduce. In 2005, Nutch used GFS and MapReduce to perform operations.

Hadoop was created by Doug Cutting and Mike Cafarella in 2005. It was originally developed to support distribution for the Nutch search engine project. Doug, who was working at Yahoo! at the time and is now Chief Architect of Cloudera.
The name Hadoop came from his son’s toy elephant. Cutting’s son was 2 years old at the time and just beginning to talk. He called his beloved stuffed yellow elephant “Hadoop” (with the stress on the first syllable). Now 12, Doug’s son often exclaims, “Why don’t you say my name, and why don’t I get royalties? I deserve to be famous for this!”
There are some similar stories about the name.

Later in Jan 2008, Yahoo released Hadoop as an open source project to Apache Software Foundation. In July 2008, Apache tested a 4000 node cluster with Hadoop successfully. In 2009, Hadoop successfully sorted a petabyte of data in less than 17 hours to handle billions of searches and indexing millions of web pages. Moving ahead in Dec 2011, Apache Hadoop released version 1.0. Later in Aug 2013, Version 2.0.6 was available, in Sep 2016, Version 3.0.0-alpha was available and in Dec 2017, Version 3.0.0 was available.

Is Hadoop dying?

Nowadays a lot of people start to talk about Hadoop is dying and that Spark is the future. But what exactly this means?
Hadoop itself is not dying but MapReduce that is batch orientate is being replaced to Spark because Spark can run in memory and with this be faster. The other thing is about the rising of the clouds, and is now possible to use cloud storage to replace HDFS and is totally possible to use tools like Spark without Hadoop. In the other hand, Hadoop 3 is supporting integration with Object Storage System and already changes yarn to run with GPU.

Hadoop Slowdown

Hadoop 3

With Hadoop 3.0 YARN will enable running all types of clusters and mix CPU and GPU intensive processes. For example, by integrating TensorFlow with YARN an end-user can seamlessly manage resources across multiple Deep Learning applications and other workloads, such as Spark, Hive or MapReduce.

The major changes are:

  • Minimum Required Java Version in Hadoop 3 is 8
  • Support for Erasure Encoding in HDFS
  • YARN Timeline Service v.2
  • Shell Script Rewrite
  • Shaded Client Jars
  • Support for Opportunistic Containers
  • MapReduce Task-Level Native Optimization
  • Support for More than 2 NameNodes
  • Default Ports of Multiple Services have been Changed
  • Support for Filesystem Connector
  • Intra-DataNode Balancer
  • Reworked Daemon and Task Heap Management


4. Courses

5. Books

Hadoop: The Definitive Guide is the best option to start.

15+ Great Books for Hadoop

6. Influencers List

7. Podcasts

8. Newsletters

9. Links

Hadoop Docs

Hadoop Architecture Overview

Hadoop complete Overview

What is MapReduce?

What is MapReduce?

MapReduce in easy way

MapReduce Concepts

Understanding Hadoop