Backpressure Flow Control
Offered By: USENIX via YouTube
Course Description
Overview
Explore a groundbreaking approach to data center network congestion control in this 16-minute conference talk from NSDI '22. Dive into Backpressure Flow Control (BFC), a novel architecture that addresses the limitations of end-to-end congestion control algorithms. Learn how BFC implements per-hop per-flow flow control with bounded state and constant-time switch operations, achieving near-optimal throughput and tail latency even under challenging network conditions. Discover the implementation of BFC on Tofino2, a P4-based programmable hardware switch, and examine simulation results showcasing significant improvements in tail latency for short flows and average completion time for long flows compared to existing schemes. Gain insights into the future of efficient data center network management and its potential impact on network utilization and operational costs.
Syllabus
Intro
Lots of Existing Congestion Control Prot . Congestion control goals
E2E Feedback Loops are too slow for • High feedback delay: network round-trip-time (E2E RTT) Acting on stale information can hurt performance
Link Speed | Uncontrolled Traffic • No feedback in the first RTT
Challenges in per-hop per-flow Flow Co
Backpressure Flow Control (BFC)
Only Track Active Flows • Active flow: flow with packets queued at the sw
Dynamic Queue Assignment
Communicate State across Swita
Evaluation: Simulation (Incast)
Thank You • Per-hop per-flow flow control is great
Taught by
USENIX
Related Courses
Scaling Memcache at FacebookUSENIX via YouTube Multi-Person Localization via RF Body Reflections
USENIX via YouTube Opaque - An Oblivious and Encrypted Distributed Analytics Platform
USENIX via YouTube Live Video Analytics at Scale with Approximation and Delay-Tolerance
USENIX via YouTube Clipper - A Low-Latency Online Prediction Serving System
USENIX via YouTube