BARSiC — Asynchronous Replication and Consensus for 70 Databases
summary
Describes BARSiC, an asynchronous replication and consensus solution built for VKontakte’s infrastructure of 70 specialized database engines across ~800 clusters and 10,000+ shards. Explains how BARSiC handles master-shard failures by providing asynchronous replication and consensus mechanisms tailored to that large, heterogeneous database landscape.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:07Current replication system
- 7:22Traditional replication issues
- 9:07Building BARSiC
- 14:05Why off-the-shelf consensus failed
- 14:49EM replication algorithm choice
- 15:56Distributed system testing approaches
- 18:03TLA+ testing model
- 25:56Testing and optimization results
- 26:41BARSiC deployment outcomes
- 27:23Q&A
related talks
38mDisaster Recovery with a Distributed Database
46mDatabase Sharding the Right Way Easy, Reliable, and Open source
42mThe hows and whys of a distributed SQL database
47mDatabase Resiliency
50mAerospike Technical Overview
24mWhat a Modern Database Enables
41mEverything you always wanted to know about highly available distributed databases
28mAn introduction to Riak