Rock around the Clock (Synchronization): Improve Performance...
summary
Demo shows a Kubernetes e‑commerce site whose tail latency jumps from ~140 ms to ~5 s when a data “elephant” workload shares the link, with system delays rising to ~1.8 ms and ~28,000 packets/sec dropped. Explains why NTP’s round‑trip assumption fails for one‑way delay visibility, compares PTP capable switches and software Hens probes that classify pure samples, and cites congestion control that pauses senders by microseconds to avoid retransmits and reduce tail latency.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 0:17Clock sync in apps matters
- 1:00Network contention demo
- 4:55Quartz and atomic clocks
- 5:58NTP and bank heist
- 8:49Precision Time Protocol
- 9:51Huygens high-precision sync
- 12:20Congestion control with sync
- 15:15Summary
related talks
26mTuning Your Cloud: Improve Global Network Performance for Applications
26mTuning Your Cloud: Improve Global Network Performance for Application
43mDo the Time Warp – the Rocky Horror PTP Show: Verification of Network Time...
26mExploiting a Natural Network Effect for Scalable, Fine-grained Clock Synchronization
24mRunning PTP at scale
45mReducing Network Latency: Innovations for a Faster Internet
25mScaling Up, One Network Bottleneck at a Time
13mLive in the Express Lane
