Mail.ru Group's own database is already 12 years old! In this post, I will share the team's successes. Over the past year, we have grown to 70 people, have done synchronous replication, managed the cluster, and got cool new customers. There are many of us, because there is a lot of interest in Tarantool in Russia. If you pay for something in our country, it is highly likely that your data goes through our database.
Tarantool stores and processes data for large projects:
- How we made the core of Alfa-Bank's investment business based on Tarantool
- Smart cache service based on ZeroMQ and Tarantool
- Tarantool at Aeroflot, or MDM on the fly
- Tarantool: history of accelerating search in 1C
Tarantool is a unique product in the database market. We are developing it together with the community and clients. This year we were ordered to develop synchronous replication. But the best part is that we were allowed to publish the result in the public domain! Synchronous replication will help increase Tarantool's market share, and now it can be used as the main database where MySQL or Postgres used to be.
Another thing that we are working on and that everyone will benefit from is working on sharding. We have accelerated the vshard module , supported and tested resharding in battle, and are making a standard API for fetching data from a cluster.
Also, now we are actively selling licenses and technical support. Along with entering the enterprise, we have a licensing policy. We also learned how to work with integrator partners. At the same time, open source remains an equally important part of our work. To keep the focus, a team of product managers emerged. As a result, we have released 2 products:
- Cartridge is an open source tool for managing Tarantool clusters with sharding, GUI and admin tools;
- Data Grid is an enterprise product for simplified work with Tarantool, which includes database schema management tools, a ready-made GraphQL API for data access and hot code deployment.
In addition, our pre-sale team has improved the skill of squeezing performance out of Tarantool to the last drops. For example, we piloted an analogue of Kafka from LinkedIn with secondary indexes and a throughput of 400 thousand requests per second and a latency of less than 5 milliseconds.
We also launched courses on all aspects of Tarantool from development to operation. We read them to our clients and teach them to use services and their resources more efficiently. We also do not forget to tell others what we are doing: we write a lot on Habr and simplify our site tarantool.io .
The most difficult thing for us is to explain to our customers why they need Tarantool. We also learned to do this better with the help of clear materials and ready-made presentations.
Tarantool has ambitious goals, so stay tuned for our news, we will certainly fulfill them and share our experience.