Explore best practices for TiKV, an open-source distributed transactional key-value database, in this 40-minute conference talk by James Zhang from PingCAP. Gain insights into controlling data distribution, recommended deployment in cross-DC scenarios, and managing data balancing during scale-out and scale-in operations. Learn to identify and address hotspot issues, fine-tune performance for various workloads, and understand TiKV's architecture, topology, and ecosystem. Discover techniques for fighting hotspots through good design, manual balancing, and auto-balancing based on statistics. Delve into performance tuning strategies, including finding bottlenecks and optimizing read operations in RocksDB.
TiKV Best Practices - Deployment, Scaling, and Performance Optimization
Overview
Syllabus
Intro
TIKV Overview and History
Architecture
st Part: Theories
1 The topology of a TIKV cluster
3.1 Scale - Transfer leader
Ecosystem
2rd Part: Practices
1.2 Deployment - Cross IDC
1.3 Demployment - Configurations
2.1 Scale - Add new nodes
2.2 Scale - Remove old nodes
3.1 Fight with hotspot
3.2 Fight with hotspots - Good design
3.4 Fight with hotspots - Find write hotspot
3.5 Fight with hotspot - Find read hotspot
3.6 Fight with hotspot - Auto balancing based on statistics
3.7 Fight with hotspot - Manual balance
4.1 Performance tuning - Find bottlenecks
4.3 Performance tuning . Read in RocksDB
Taught by
Linux Foundation
