DISC: Backpressure Mitigation In Multi-tier Applications With Distributed Shared...
summary
DISC lets multiple tiers jointly act as endpoints of the shared TCP connection so backend servers can emit large response bodies directly to clients while headers and trailers traverse the normal path. Implementation uses kernel BPF for redirection, TCP sequence-number translation, and distributed TLS sessions, requires only small localized code changes, supports arbitrary multi‑tier topologies and heterogeneous protocols, and reports up to 41.5% CPU reduction, 74.1% average latency drop, 5.71× tail improvement, and 45% higher request rate.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:21Multi-tier app example
- 3:56Problem with CPU load
- 5:26Backpressure limits
- 8:44DISC protocol design
- 10:11BPF and TLS implementation
- 11:15DISC workflow
- 14:08Results and scalability
- 14:52Conclusion
related talks
29mFault Tolerance is a Requirement Not a Feature - Adar Danait, Lilian Ernest, Citi
22mBirds of a Feather - xDS and Load Balancing
44mDeep Dive talk | Resilience in Microservices with Congestion Control
21mHigh-Performance Cloud Native Traffic Authentication Solutions
44mArchitectural Patterns of Resilient Distributed Systems
38mTechniques and Tools for a Coherent Discussion about Performance in Complex...
43mDistributed Systems
39mWhen "Worst" is Best (in Distributed Systems)