Scaling Limits of the Sokol Disk DBMS
summary
Examines scaling limits of a disk-based DBMS called Sokol, evaluating whether a pure disk-only database (without caching front layers) can meet practical requirements for simplicity, reliability and lower hardware needs. Discusses which DBMS designs satisfy known scalability constraints and analyzes trade-offs when using a disk-only approach.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:10Sokol DBMS architecture
- 2:35Connection handling resources
- 5:54Group commits performance
- 9:31Scaling synchronization
- 12:44Caching I/O
- 13:53Disk benchmarks
- 16:31Comparison in-memory systems
- 20:50Hardware limits
- 24:49Scaling per cores
- 31:18Amdahl's law conclusions
- 33:26Q&A
related talks
30mDatastore Axes: Choosing the Scalability Direction You Need
34mWhat We Talk About When We Talk About Disk IO
12mSpanDB: A Fast, Cost-Effective LSM-tree Based KV Store on Hybrid Storage
27mScaling for Intelligence: Distributed SQL Databases in the Age of AI
30mOvercoming Distributed Databases Scaling Challenges with Tablets
14mRe-think Data Management Software Design Upon the Arrival of Storage Hardware with
49mSSDs for Your Database
29mModern Database Architecture: From SQL to NoSQL & Distributed SQL