How do you compare to Yandex YDB ?

[samber]

Yandex released a distributed transactional database recently.

What are your strengths compared to YDB?

[jordanlewis]

I haven't spent time analyzing YDB, so this is just an initial surface-level opinion based on reading the docs.

CockroachDB supports standard, PostgreSQL-compatible SQL and is compatible with PostgreSQL drivers, wire protocol, and tools. This means that you can migrate existing applications that use PostgreSQL to using CockroachDB without much trouble. On the other hand, based on my cursory examination of YDB, YDB uses a custom SQL dialect that is specific to YDB, which requires using YDB SDKs, YDB syntax, YDB DDL methods, and so on.

CockroachDB has a very strong observability story, with pages and metrics that allow you to deeply understand the performance of your workload.

CockroachDB has a state-of-the-art cost-based optimizer, a vectorized SQL execution model, and a distributed SQL execution runner that allows it to execute complex HTAP workloads with good efficiency and performance. I'm unsure how YDB performs on this front. And likewise for the transaction model - YDB appears to use only optimistic concurrency control, whereas CockroachDB uses a blend of optimistic and pessimistic based on our years of experience running the database at scale and finding interesting challenges with a purely optimistic concurrency control model.

I recommend reading CockroachDB's SIGMOD paper to learn more about its transaction model and more: https://dl.acm.org/doi/pdf/10.1145/3318464.3386134