Distributed SQL How TiDB Modernized Data Management
summary
TiDB is a MySQL‑compatible distributed SQL system that uses TiKV + Raft, PD (placement driver) and a strongly consistent column store TFlash to avoid manual sharding and give ACID at scale. It supports TiDB Serverless with S3 as canonical storage and local NVMe/EBS caching via a custom Cloud Engine to cut costs and move compaction to spot nodes. Production shown numbers include 100 billion rows/month and 1M read‑writes/sec; online DDL can build a 10 TB index in ~40 minutes. Resource groups use mclock scheduling to protect short queries.
AI-written summary · about
chapters (19)
- 0:00Introduction to TiDB
- 1:06History and architecture evolution
- 3:14Spanner vs TiDB design
- 4:41TiDB’s design philosophy
- 7:10TiKV and distributed storage
- 8:33PD as the cluster brain
- 10:16TiFlash column store
- 13:51TiDB’s multi-region scalability
- 14:14Customer benefits and use cases
- 15:18Why users switch to TiDB
- 16:20TiDB Serverless goals
- 18:06Multi-tenancy requirements
- 18:48TiDB Serverless architecture
- 20:57Cloud Engine storage engine
- 22:26Leveraging S3 for durability
- 23:55Balancing storage costs
- 25:46Resource control in TiDB
- 27:10Resource groups for workloads
- 28:11Scheduling and consistency
related talks
47m10 Reasons Why the TiDB Database Should Power Your Next Development Project
55mThe Future of MySQL
33mProvisioning the Distributed Database TiDB on Kubernetes
49mThe Paradigm Shift - Understanding the Move To Distributed Databases
24mUsing a Distributed Key-Value Store
46mTiDB 101
18mTiFlash: Make TiKV 10x Faster and HTAP-able
29mHow We Make TiKV- a Distributed Storage, More Cost-Effective On the Cloud