The Paradigm Shift - Understanding the Move To Distributed Databases
summary
Claim: a distributed SQL (PingCAP TiDB) implements strong ACID across ranges using Raft replication plus a decentralized two‑phase commit, stores tables and secondary indexes as key‑value pairs in TiKV, splits data into ~96 MB regions, supports MVCC, HTAP with synced column/row stores, nonblocking schema changes and online scale‑out. Demo shows a local playground cluster, live inserts during index build, automatic region rebalancing and no application downtime.
AI-written summary · about
chapters (21)
- 0:00Start
- 2:32New Applications Requirements
- 3:31Examples of New Applications
- 4:32Build an E-Commerce Platform
- 5:21Characteristics of E-Commerce Platform
- 6:02When the Scale is Small
- 6:48The Happy Problem: Scalability • Number of users increases
- 8:53Issues with Sharding • Application development
- 11:16Issues with NoSQL
- 13:51Issues with Complex Architecture • Complex
- 16:46Scaling E-Commerce Platform is not Alone • Complex tech stack is used for new applications • Application developers are solving database problems
- 18:30What's on the Wishlist for a Database?
- 20:34TiDB Architecture
- 21:45Data Organization
- 27:48Parallel Query Execution
- 29:15HTAP
- 32:32High Availability
- 33:06Timeliness
- 34:06Reference Architecture
- 35:06Summary
- 35:49Demo
related talks
29mDistributed SQL How TiDB Modernized Data Management
29mDistributed SQL: How TiDB Modernized Data Management
24mUsing a Distributed Key-Value Store
40mTiKV Best Practices
29mModern Database Architecture: From SQL to NoSQL & Distributed SQL
31mSharded & Distributed Are Not the Same:
28mHow Does Geo-Replication Work in TiDB?
42mThe hows and whys of a distributed SQL database