PingCAP_A TiDB Story CTO技术分享-大数据.pdf

How can HTAP help you A TiDB Story Who am I - Dongxu Huang - CTO & Cofounder, PingCAP - Distributed system engin r / Open source advocator - h@ - TiDB / TiKV / TiSpark But wait... - How to scale? - Eventual consistency can confuse developers and applications - How to analyze the data in it? - How to integrate with cloud infrastructur And ... A bit of history... - Traditional relational database is not designed for distributed environment - Even if you shard it out, you’re basically replicating the same stack What changed since 2000s... - Hardware is becoming more and more powerful, and cheaper - Distributed system is the mainstream - Distributed storage stack is highly layered nowadays Our Goal - Full-featured SQL - With MySQL syntax and wire protocol - ACID Transaction - Push-button scaling - HA without pain - Can serve both OLTP & OLAP workload - No more ETL - Open-source, of course. Use case 1: MySQL shard consolidation * What’s Syncer? Syncer Use case 2: Drop-in replacement for MySQL Application Application Application - Massive and concurrent workload - OLTP on a large dataset - TiDB will balance the workload automatically, and it’s transparent to the application layer - Sysbench result TiDB Use case 3: Real-time analytics and reporting Use case 3: Real-time analytics and reporting Use case 4: Building block for other systems TiKV Logical View YCSB Benchmark (2