Graduated Project Lightning Talk: Time to Live (TTL) Support for TiKV
summary
TiKV adds per-key TTL by appending an 8-byte expiry timestamp to each value, checking that expiry on reads and optionally using the PD global timestamp for linearizability; this avoids RocksDB’s single-TTL limitation. Disk reclamation uses RocksDB compaction filter drops expired entries plus a user-collected SST max-expiry property so a background worker can call CompactFiles on SSTs whose max expiry is past due. Project is graduated with ~9,000 stars, ~300 contributors and ~1,500 production adopters, and the approach reclaims space without extra reads/writes.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 0:38What is TiKV
- 1:26TiKV Architecture
- 3:23What is TTL
- 4:08Limitations of TTL
- 5:09TTL in TiKV
- 6:06Linearizability
- 9:23Outro
related talks
29mEfficient and Consistent Time to Live TTL Implementation in MyRocks at Meta
40mTiKV Best Practices
11mAdapting TiKV for Cloud Storage
40mTiKV Best Practices
18mTiFlash: Make TiKV 10x Faster and HTAP-able
29mHow We Make TiKV- a Distributed Storage, More Cost-Effective On the Cloud
47m10 Reasons Why the TiDB Database Should Power Your Next Development Project
18mHow We Reduced Write Latency in TiKV?