Rearchitecting Linux Storage Stack for µs Latency and High Throughput

Explore a groundbreaking conference talk that challenges the widespread belief about Linux's inability to achieve microsecond-scale latency and high throughput simultaneously. Dive into the innovative blk-switch architecture, a new Linux kernel storage stack design that adapts techniques from computer networking to revolutionize storage performance. Learn how this approach achieves microsecond-scale average and tail latency while allowing applications to fully utilize hardware capacity, even in high-contention scenarios. Discover the key insights behind blk-switch, including its conceptual similarity to network switches, prioritized processing, request steering, and application steering. Examine the evaluation setup, performance breakdown, and impressive results that demonstrate the potential for microsecond-scale latency in Linux, all without modifying applications, network hardware, or kernel components.

Intro
Widespread belief: Linux can't achieve ps-scale latency & high throughput
Performance of Existing Storage Stacks Applications accessing in-memory data in remote servers (single-core case)
blk-switch Summary
blk-switch Key Insight • Observation: Today's Linux storage stack is conceptually similar to network switches!
A deeper dive into blk-switch architecture
blk-switch Prioritization
blk-switch Request Steering for transient loads
blk-switch Application Steering for persistent loads
blk-switch Evaluation Setup • Implemented entirely in the Linux kernel with minimal changes (LOC-928)
High Contention Scenario (In-memory)
blk-switch Performance Breakdown
Summary . It is possible to achieve pes-scale latency and high throughput with Linux