Adapting TiKV for Cloud Storage
summary
TiKV adds Raft Engine — an LZ4‑compressed, append‑only log with an in‑memory index that replaces RocksDB and cuts background and server writes by about 30% — and a Prioritized I/O Rate Limiting scheme that tags IO as A/B/C and throttles lower priorities when global IO exceeds a predetermined limit, implemented without new IO queues or task rewrites and shown to stabilize throughput during large imports. Experiments also include CPU limiting for low‑core nodes and a raft witness write‑only replica to lower replication cost.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 0:23TiKV Overview
- 1:07Hardware Design
- 2:22Scaling
- 4:30Raft Engine
- 5:31Graph Engine
- 6:52Priority IO Scheduling
- 9:01Demonstration
- 9:31Conclusion
related talks
29mHow We Make TiKV- a Distributed Storage, More Cost-Effective On the Cloud
40mTiKV Best Practices
18mHow We Reduced Write Latency in TiKV?
24mUsing a Distributed Key-Value Store
40mTiKV Best Practices
18mTiFlash: Make TiKV 10x Faster and HTAP-able
29mDistributed SQL How TiDB Modernized Data Management
28mHow Does Geo-Replication Work in TiDB?