Cache? Where we're going, we don't need cache!
summary
Claims that ScyllaDB can replace an external cache with its embedded read‑through cache and hit <1 ms P99 for some workloads backed by customer examples. External cache problems include higher overall latency from unexpected cache misses, doubled operational cost and loss of DB query metadata. Migrations from Redis + Elasticsearch by Disney Plus Hotstar and from Redis by Grab report lower cost and stable performance, and ScyllaDB exposes controls like bypass cache to tune per query behavior.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:59Low latency in databases
- 3:00Database latency explained
- 7:56Tail latency impact
- 11:33Latency with cache
- 13:39Caching strategies trade-offs
- 16:24External cache problems
- 19:52ScyllaDB as cache
- 22:43ScyllaDB cache features
- 25:11Q&A
related talks
41mPerformance & Fault Tolerance: Building a Modern Database
44mScyllaDB where Cassandra meets Redis!
01h 02mBreak Out of Redis
35mHow to Save on Scaling by Moving from Cassandra to Scylla DB
31mSolid Cache: A disk backed Rails cache
26mDesigning a Query Queue for ScyllaDB
15mAccelerate Application Response Times with True Cache
44mCan we go with a database but without a cache layer in 2024?