Using a Distributed Key-Value Store
summary
TiKV is a distributed, transactional key‑value store that shards data into regions, replicates with Raft, and uses a PD timestamp oracle for transactions; customers run clusters from 200-node cluster with 100M reads/sec on 2T rows (~320 TB) up to ~1600 nodes. Clients talk over gRPC and choose raw vs transactional APIs (raw = single‑row, low latency; txn = MVCC, snapshot isolation). Recent protocol changes cut read/write latency, TiKV adds compare-and-swap and a pluggable coprocessor to run computation near data.
AI-written summary · about
chapters (6)
- 0:00Start
- 5:45Hbase
- 6:29Mongodb
- 7:46Taikov
- 13:21Architecture of a Thai Kv Cluster
- 18:48Using the Python Client
related talks
40mTiKV Best Practices
11mAdapting TiKV for Cloud Storage
29mDistributed SQL How TiDB Modernized Data Management
40mTiKV Best Practices
29mDistributed SQL: How TiDB Modernized Data Management
29mHow We Make TiKV- a Distributed Storage, More Cost-Effective On the Cloud
49mThe Paradigm Shift - Understanding the Move To Distributed Databases
18mTiFlash: Make TiKV 10x Faster and HTAP-able