Overview
Explore the future of FoundationDB storage engines in this 26-minute conference talk by Steve Atherton from Apple. Delve into the role of FDB's storage engine component and how its requirements are shaped by FDB's distributed architecture. Gain insights into the upcoming storage engine specifically designed for FDB's read and commit paths. Learn about its advantages, including higher write throughput, smaller space footprint through key prefix compression, and configurable support for long-running read transactions. Discover the trade-offs between disk space and data version persistence. Examine the FoundationDB SUMMIT overview, FDB storage architecture, current storage engines, and the motivation behind new developments. Investigate the Multi-Version Layer, 'missing' requirements, and the 'ssd' storage engine. Compare B+Tree and LSM structures before diving into the Redwood Storage Engine, featuring Copy-On-Write B+Tree, versioned pager, and versioned B+Tree concepts. Explore non-linear prefix compression techniques and their application in binary and B+Trees.
Syllabus
FoundationDB SUMMIT
Overview
FDB Storage Architecture
Current Storage Engines
Motivation
Multi-Version Layer
Missing' Requirements
More About 'ssd' Storage Engine
B+Tree vs LSM
Redwood Storage Engine
Copy-On-Write B+Tree Intro
Limit Cow using Indirection
Redwood Versioned Pager
Redwood Versioned B+Tree
Non-Linear Prefix Compression
Prefix Compressed Binary Tree
Prefix Compressed B+Tree
Conclusion
Taught by
Linux Foundation