Data Engineering *

I had some experience in the matching engine development for cryptocurrency exchange some time ago. That was an interesting and challenging experience. I developed it in clear C++ from scratch. The testing of it is also quite a challenging task. You need to get data for testing, perform testing, collect some statistics, and at last, analyze collected data to find weak points and bottlenecks. I want to focus on testing the C++ matching engine and show how testing can give insights for optimizations even without the need to change the code. The matching engine I developed can do more than 1’000’000 TPS (transactions per second) and is 10x times faster than the matching engine of the Binance cryptocurrency exchange (see one post on Binance Blog).

Recently we released a new build of the Big Data Tools plugin that is compatible with the 2021.3 versions of IntelliJ IDEA and PyCharm. DataGrip 2021.3 support will be available immediately after the release in October. The plugin also supports our new data science IDE – JetBrains DataSpell. If you still use previous versions, now is the perfect time to upgrade both your IDE and the plugin. 

This year, we introduced a number of new features as well as some features that have been there for a while, for example, running Spark Submit with a run configuration.

Here’s a list of the key improvements:

In this topic, I will tell you how to dynamically parse and deserialize only part of the whole JSON document. We will create an implementation for .NET Core with C# as a language.

For example, we have the next JSON as a data source for the report. Notice that we will get this JSON in the runtime and at the compile step we don't know the structure of this document. And what if you need to select only several fields for processing?

Hey, hey! I am Ilya Kalchenko, a Data Engineer at NIX, a fan of big and small data processing, and Python. In this article, I want to discuss the benefits of hybrid data lakes for efficient and secure data organization.

 To begin with, I invite you to figure out the concepts of Data Warehouses and Data Lake. Let’s delve into the use cases and delimit areas of responsibility.

Home of this article: https://robotics.snowcron.com/coins/02_head_or_tail.htm

The global objective of these articles is to build a coin classifier, capable of scanning your pocket change and find rare / valuable coins. This is a second article in a series, so let me remind you what happened earlier (https://habr.com/ru/post/538958/).

During previous step we got a rather large dataset composed of pairs of images, loaded from an online coins site meshok.ru. Those images were uploaded to the Internet by people we do not know, and though they are supposed to contain coin's head in one image and tail in the other, we can not rule out a situation when we have two heads and no tail and vice versa. Also at the moment we have no idea which image contains head and which contains tail: this might be important when we feed data to our final classifier.

So let's write a program to distinguish heads from tails. It is a rather simple task, involving a convolutional neural network that is using transfer learning.

Same way as before, we are going to use Google Colab environment, taking the advantage of a free video card they grant us an access to. We will store data on a Google Drive, so first thing we need is to allow Colab to access the Drive:

See more at robotics.snowcron.comThis is the first article in a serie dedicated to coins classification.Having countless "dogs vs cats" or "find a pedestrian on the street" classifiers all over the Internet, coins classification doesn't look like a difficult task. At first. Unfortunately, it is degree of magnitude harder - a formidable challenge indeed. You can easily tell heads of tails? Great. Can you figure out if the number is 1 mm shifted to the left? See, from classifier's view it is still the same head... while it can make a difference between a common coin priced according to the number on it and a rare one, 1000 times more expensive.Of course, we can do what we usually do in image classification: provide 10,000 sample images... No, wait, we can not. Some types of coins are rare indeed - you need to sort through a BASKET (10 liters) of coins to find one. Easy arithmetics suggests that to get 10000 images of DIFFERENT coins you will need 10,000 baskets of coins to start with. Well, and unlimited time.So it is not that easy.Anyway, we are going to begin with getting large number of images and work from there. We will use Russian coins as an example, as Russia had money reform in 1994 and so the number of coins one can expect to find in the pocket is limited. Unlike USA with its 200 years of monetary history. And yes, we are ONLY going to focus on current coins: the ultimate goal of our work is to write a program for smartphone to classify coins you have received in a grocery store as a change.Which makes things even worse, as we can not count on good lighting and quality cameras anymore. But we'll still try.In addition to "only Russian coins, beginning from 1994", we are going to add an extra limitation: no special occasion coins. Those coins look distinctive, so anyone can figure that this coin is special. We focus on REGULAR coins. Which limits their number severely.Don't take me wrong: if we need to apply the same approach to a full list of coins... it will work. But I got 15 GB of images for that limited set, can you imagine how large the complete set will be?!To get images, I am going to scan one of the largest Russian coins site "meshok.ru".This site allows buyers and sellers to find each other; sellers can upload images... just what we need. Unfortunately, a business-oriented seller can easily upload his 1 rouble image to 1, 2, 5, 10 roubles topics, just to increase the exposure.

So we can not count on the topic name, we have to determine what coin is on the photo ourselves.To scan the site, a simple scanner was written, based on the Python's Beautiful Soup library. In just few hours I got over 50,000 photos. Not a lot by Machine Learning standards, but definitely a start.After we got the images, we have to - unfortunately - revisit them by hand, looking for images we do not want in our training set, or for images that should be edited somehow. For example, someone could have uploaded a photo of his cat. We don't need a cat in our dataset.First, we delete all images, that can not be split to head/hail.

Challenges of real-time AI/ML computations

• A “high-load” customer portal is integrated with an online recommendation system. The plan is to reconfigure promo campaigns at the level of the entire retail network (we will assume that instead of a “flat” promo campaign master there will be used a “segment-tactic” matrix). What will happen to the recommender mechanisms? What will happen to data feeds and updates into the recommender mechanisms (the volume of input data having increased 25000 times)? What will happen to recommendation rule generation setup (the need to reduce 1000 times the recommendation rule filtering threshold due to a thousandfold increase of the volume and “assortment” of the rules generated)?
• An equipment health monitoring system uses “manual” data sample feeds. Now it is connected to a SCADA system that transmits thousands of process parameter readings each second. What will happen to the monitoring system (will it be able to handle equipment health monitoring on a second-by-second basis)? What will happen once the input data receives a new bloc of several hundreds of columns with data sensor readings recently implemented in the SCADA system (will it be necessary, and for how long, to shut down the monitoring system to integrate the new sensor data in the analysis)?
• A complex of AI/ML mechanisms (recommendation, monitoring, forecasting) depend on each other’s results. How many man-hours will it take every month to adapt those AI/ML mechanisms’ functioning to changes in the input data? What is the overall “delay” in supporting business decision making by the AI/ML mechanisms (the refresh frequency of supporting information against the feed frequency of new input data)?

What do these terms mean? And what is the difference?

Dear all,

I am the head of Data Science at British Transport Police, and one of our department tasks is to efficiently allocate staff, depending on the crime rates, which correlate to passenger flow. As you understand, the passenger flow will undertake significant change as soon as the Government decides to cancel quarantine or stop some limitations. The question naturally arises: when will the pandemic end and how to prepare for a return to normal life.

If you utilize Apache Spark, you probably have a few applications that consume some data from external sources and produce some intermediate result, that is about to be consumed by some applications further down the processing chain, and so on until you get a final result.

We suspect that because we have a similar pipeline with lots of processes like this one:

Click here for a bit larger version

Each rectangle is a Spark application with a set of their own execution parameters, and each arrow is an equally parametrized dataset (externally stored highlighted with a color; note the number of intermediate ones). This example is not the most complex of our processes, it’s fairly a simple one. And we don’t assemble such workflows manually, we generate them from Process Templates (outlined as groups on this flowchart).

So here comes the One Ring, a Spark pipelining framework with very robust configuration abilities, which makes it easier to compose and execute a most complex Process as a single large Spark job.

And we just made it open source. Perhaps, you’re interested in the details.