Compaction for All! Unified Compaction Strategy in Cassandra 5.0 (CEP-26)
summary
Cassandra 5.0 introduces Unified Compaction Strategy (UCS) that replaces LCS/STCS with a single density-based sharding approach: SSTable density (size/token fraction) drives level promotions, shards and compaction selection, producing examples like 430 Mebit overlap becoming eight shards and two 135MB + two 80MB outputs. UCS is configurable per-level (w values map to legacy L10 or STCS), parallelizes compaction across shard token ranges, is stateless so a strategy change avoids a full rewrite, and reduces space, read and write amplification in practice.
AI-written summary · about
chapters (8)
- 0:00Introduction to compaction
- 1:23Challenges in data storage
- 4:15Legacy compaction strategies
- 8:37Unified compaction strategy
- 12:21Deep dive into UCS mechanics
- 21:10Sharding and performance
- 27:25Example of compaction steps
- 31:44Comparison and conclusion
related talks
29mThe Unified Compaction Strategy in Cassandra 5
01h 02mBuilding state-of-art LSM-tree compaction in ScyllaDB
45mCassandra Architecture: Understanding, Scaling and Optimizing
20mCastle-enhanced Cassandra
54mEvolving Transactions in Apache Cassandra
24mTSDB Data Compactions on Scale Without Headache: Now & Future
29mCassandra - an introduction
53mApache Cassandra. Inception - all you need to know