SyncWise: Error-Aware Time Synchronization for Reconfigurable Data Center Networks
summary
SyncWise is a synchronization protocol for reconfigurable data center networks that predicts per-node error bounds from preprofiled drift and hop-error limits and only syncs to neighbors that reduce that bound, yielding provably optimal parent selection. It is prototyped on a testbed with an OCS and Intel Tofino2 switches and an open‑source event‑driven simulator. Results show sub-10 ns max sync error, beating Firefly by 2.5× and Graham by 3.8× and enabling guard bands sized as 2× worst‑case sync error plus OCS switch time.
AI-written summary · about
chapters (8)
- 0:00SyncWise introduction
- 1:44Time sync importance
- 4:32Existing sync limitations
- 5:37Predicting sync error bounds
- 8:27Error-aware protocol design
- 11:38SyncWise implementation
- 12:43Key results comparison
- 13:25Summary
related talks
26mExploiting a Natural Network Effect for Scalable, Fine-grained Clock Synchronization
28mRock around the clock (synchronization
12mDebugging Transient Faults in Data Centers using Synchronized Network-wide Packet
15mParaSync: Exploiting Fine-Grained Parallelism for Efficient File Synchronization
14mOrbWeaver: Using IDLE Cycles in Programmable Networks for Opportunistic Coordination
14mLightning Talk: Beyond NTP: High-accuracy clocks for distributed observability
12mSyncScatter: Enabling WiFi like synchronization and range for WiFi backscatter
21mTime Synchronization using ML Techniques in Electronic Trading