Anti-Patterns with Apache Cassandra
summary
Cassandra clusters suffer two concrete anti-patterns: creating huge numbers of tables and growing oversized partitions. Table metadata consumes about ~1 MB heap per table, many small memtables trigger shared flush bottlenecks (default 2 flush writer threads), and load tests show p999 latency falls as table count drops; recommendation is to keep tables under 200. Partitions should stay under ~10 MB, avoid collections >64 KB, offload large payloads to object or memory stores, and validate with cassandra-stress, noSQLBench or Locust and virtual table stats.
AI-written summary · about
chapters (9)
- 0:00Cassandra anti-patterns intro
- 1:47Performance issues in Cassandra
- 3:53Too many tables anti-pattern
- 8:27How many tables are too many
- 12:20Designing with fewer tables
- 12:59Large partitions anti-pattern
- 16:30Case studies: large partitions
- 18:59Partition and table design tips
- 21:05Q&A
related talks
23mDeciphering Cassandra’s Prophecy: Preventing the Fall of Your Application
53mApache Cassandra. Inception - all you need to know
29mCassandra - an introduction
31mCassandra for giant 3D tables
29mTime Series with Apache Cassandra: the basics
46mCassandra Memory Explained. Cassandra memory footprint
57mCassandra for Metadata Storage: Successes and Failures
45mCassandra Architecture: Understanding, Scaling and Optimizing