Introduction to YugabyteDB: Design and Architecture
summary
YugabyteDB is a distributed SQL database that shards tables into tablets, replicates each tablet with a typical replication factor three, and uses Raft consensus to synchronously commit writes so a majority ack guarantees durability. Distributed transactions use provisional writes and an internal two‑phase commit to provide atomicity with snapshot and serializable isolation. Storage is a document layer built on DocDB on RocksDB, optimized for SSDs; masters manage re‑replication and failover completes in seconds, while async read replicas serve low‑latency stale reads.
AI-written summary · about
chapters (19)
- 0:00Intro
- 0:21What is Distributed SQL?
- 2:35Monolithic Databases vs Distributed Databases
- 4:27How do Distributed Databases Scale Out
- 5:12How do Distributed Databases Tolerate Failures
- 6:40Network Partition
- 8:00CAP Theorem
- 10:39ACID Compliance
- 12:33Components
- 14:19Component Services
- 15:14YB-TServers
- 16:11YB-Masters
- 17:21Sharding Layer
- 18:46Replication Layer
- 20:21Storage Layer
- 22:01Secondary Indexes
- 22:58Query Layer
- 25:01Existing PostgreSQL Architecture
- 26:25Self-Healing Against Failures
related talks
27mIntroduction to YugabyteDB – Design and Architecture
25mYugabyteDB Technical Overview
47mAnatomy of a Distributed SQL Database - YugabyteDB
01h 27mYugabyteDB YSQL Development
30mTechnical Comparison of YugabyteDB with Oracle and PostgreSQL
29mEnterprise Scale: How a Distributed Architecture Can Deliver Scale and Resilience
26mIntroducing YugabyteDB for PostgreSQL Users
30mBuilding Cloud Native Applications on Distributed SQL Platform