Graham: Synchronizing Clocks by Leveraging Local Clock Properties
Talkat 19th USENIX Symposium on Networked Systems Design and Implementation (NSDI '22) · 2022-04-04..04-06·
summary
Graham is a system that characterizes each server's quartz oscillator as a temperature-to-frequency curve and uses commodity temperature sensors plus least-squares fitting over NTP to predict and correct drift. It converges in about 48 hours (NTP) to within ~0.5 ppm of GPS-derived curves, selects the best sensor per machine, and survives thermal shocks keeping errors under ~10 µs; overall it reduces assumed drift from 200 ppm to about 100 ppb, up to 2000× improvement.
AI-written summary · about
chapters (6)
related talks
26mExploiting a Natural Network Effect for Scalable, Fine-grained Clock Synchronization
30mOk, it's time to synchronize our clocks...
43mTic-Tac... Bomb! Master time and NTP before it's too late
25mRF System Synchronization - Baseband
28mRock around the clock (synchronization
14mLightning Talk: Beyond NTP: High-accuracy clocks for distributed observability
35mNobody Knows What Time It Is
43mDo the Time Warp – the Rocky Horror PTP Show: Verification of Network Time...