Report from Tarantool Meetup January 28

On January 28, Tarantool Meetup was held at the Mail.Ru Group office, where the advantages and features of Tarantool were discussed, as well as the experience of using this DBMS and its development plans. Under the cat you can find videos and presentations from these performances.

Denis Anikin, Mail.Ru Group, "Tarantool - case studies in Mail.Ru Group"

Denis told about three cases of using Tarantool in Mail.Ru Group - this is a user authentication system, a notification system for mobile applications and an advertising display system. In all three cases, Tarantool is the cornerstone of a distributed server infrastructure that serves a total of about 100 million users per month.
')
The decision to develop your own general purpose DBMS (database management system) usually takes a long time to mature. Before we rush into this abyss, we first look for whether it is possible to solve current problems with existing tools, including creating them on top of them or integrating them into various solutions. And only when we see a bright image of an ideal database, possessing all the necessary features, do we rush into all serious and begin to create.

This is exactly how Tarantool appeared - one of the fastest databases in the world, which is widely used in Mail.Ru Group and beyond. Tarantool was developed from the outset by a separate team, open source and with an eye on general use. As a result, we got an excellent product - a general purpose DBMS with a unique set of features.

Performance video



Alexander Golovko, Wallarm, "The Practice of Using Tarantool in Products: The Wallarm Experience"

The report presented the practical results of a two-year experience of using Tarantool in an alienable software solution.

We needed to perform asynchronous analytics using data from short-lived HTTP scan objects. We studied a number of possible solutions to this problem, and as a result we chose Tarantool and the concept application-server. When designing our web application security system, we were faced with the need to correlate data from several HTTP requests that were already processed. The most effective solution to this problem was the separation of query processing logic between the synchronous process of calculating parameters for analysis and the asynchronous process of the analyst itself. In fact, we brought to Tarantool all the post-processing of HTTP requests after their initial analytics on the Nginx side. As a result, it has become possible to detect various behavioral attacks on web applications, for example, brute force or code.

Alexander conducted a comparison of the characteristics of all the considered tools. There were also arguments and criteria for choosing the architecture. The final part of the presentation is devoted to the problems that developers have encountered in the implementation, operation and migration to new versions of Tarantool.

Performance video



Dmitry Novikov, Badoo, “User sessions in Badoo”

The report explained why we needed sessions, where Badoo had stored them before, that they thought up, why they decided to use Tarantool, and what all this led to.

Before we started using Tarantool, Badoo user sessions were stored in a self-written daemon. The demon was good, but there were some problems with it. For example, he stored data only in memory. And in general, everything that happened with the sessions began to cause certain inconveniences. Therefore, in order for software development to bring joy and pleasure, we decided to radically alter the system working with sessions.

Performance video