Branimir Lambov from IBM on Cassandra

a day ago

Branimir Lambov, a Cassandra committer and PMC member at IBM, discusses his career journey into Cassandra despite initially finding databases uninteresting.
He highlights key projects: deterministic token allocation, modernizing compaction strategies resulting in Unified Compaction Strategy, and integrating Trie into LSM Tree.
The Trie project began nearly a decade ago to leverage byte order in Cassandra, leading to the BTI format and trie memtables, enhancing performance and storage.
Adoption of new features like BTI and trie memtables has been significant in DataStax products, though broader community uptake in Cassandra 5 may take time.
Branimir emphasizes lessons from past incidents, including the risks of assuming code is unimportant and the pitfalls of assertions that can be disabled.
On programming languages, he acknowledges Java's drawbacks but appreciates the JVM and new features like virtual threads; he also expresses ambivalence about using English as a programming language with LLMs.
He mentions excitement for the Accord project, which will introduce ACID transactions to Cassandra with high performance.
Branimir prefers IntelliJ for its reliable tools and refactoring capabilities.

Hasty Briefsbeta