On Friday, December 9, the IT HighRise meet-up will be held in Tbilisi. It’ll be the first one in a series of events for IT developers organized by CoinsPaid and CoinsPaid Media. We talked to one of the speakers — Alexey Goncharuk, CTO of Querify Labs, who will talk about coordinating transactions in distributed databases.
— Alexey, tell us about your projects. What are the most interesting cases in your professional activities? What were the challenges and non-standard solutions?
— I’ve been working with distributed systems for more than ten years. I started to get acquainted with them in GridGain, working on a product of the same name, which later became the Apache Ignite platform. In GridGain, I worked on transaction and replication protocol improvements and managed the development of persistent storage (Apache Ignite native persistence).
Currently, I’m the CTO at Querify Labs, where we help large businesses and tech companies develop DBMS components using the experience we’ve accumulated over the years. In addition to Apache Ignite, our team has worked on Hazelcast, Yandex Database, and Clickhouse projects. We gained solid experience in developing SQL query optimizers, SQL runtimes, data storage systems, and protocols, which allows us to help our clients build custom systems tailored to their specific needs and therefore get a significant performance boost. Basically, our expertise makes it possible to build a custom DBMS from existing open-source “semi-finished products.”
— What key points will you cover in your speech at the IT HighRise event?
— At first, I’d like to talk about some widely used transaction protocols, but while preparing my presentation, I realized that it would be much more interesting and useful to discuss the problem of transaction coordination experienced by users, what fundamental limits exist with transactions, and how this can affect the architecture of the final systems.
— Where do you see data management going in 2023?
— Based on our experience at Querify Labs, I think we’ll be seeing more use of low-level frameworks such as Apache Arrow, Apache Calcite, and Facebook Velox, following the principle of composable systems — with a general trend away from off-the-shelf solutions toward custom data processing systems that will use existing building blocks for high performance and federal systems like Trino, Dremio, etc.
