Synchronous Geo-Replication over Azure Tables
summary
RTable is a client‑side library that implements chain replication on top of the Azure Table SDK to provide synchronous geo‑replication across regions; writes commit only at the tail and global writes measure about 200 ms write latency while nearest‑replica reads are ~8 ms. Demo runs across US East/West, shows live failover, adds/removes replicas without blocking reads or writes, supports consistent snapshots and cross‑table transactions, runs in production and will be released on GitHub.
AI-written summary · about
chapters (17)
- 0:00Intro
- 0:37Software
- 1:30Agenda
- 3:40Scale for Azure
- 4:59Our Table
- 9:16Key Components
- 9:29Azure Tables
- 10:52How it works
- 13:35Moving left to right
- 15:23Serializability
- 16:56Demonstration
- 21:18Recovery
- 23:42Configuration Management
- 25:02What the team learned
- 26:53Snapshots
- 27:46Primitives
- 28:41Storage Layer
related talks
26mData Replication Options for Azure SQL PaaS Services
28mHA and DR at a glance with Azure Database for PostgreSQL
31mData Replication in Azure SQL and Beyond
38mHow Geo-Distributed Apps Use Global Cloud Infra for Speed and Compliance
17mMako: Speculative Distributed Transactions with Geo-Replication
28mHow Does Geo-Replication Work in TiDB?
22mCloud-Native Apps, recovery ready!
47mPerfecting business continuity for Azure SQL DB