Analyze TCP tcp packet trace between two nodes with latency and under heavy load

celestiaorg / celestia-app

PoS application for the consensus portion of the Celestia network. Built using celestia-core (fork of CometBFT) and the cosmos-sdk

https://celestia.org

Apache License 2.0

328 stars 261 forks source link

Analyze TCP tcp packet trace between two nodes with latency and under heavy load #3613

Open evan-forbes opened 4 days ago

evan-forbes commented 4 days ago

Looking at the tcp packet traces for a local experiment showed that with latency and under heavy loads, the tcp connection was unable to utilize all of its throughput because the server (comet's p2p stack) was unable to clear it fast enough.

AC

Look for slow downs in tcp packets to determine concretely if the tcp connection is being under utilized, or if there is some form of blocking. Report back here to close.

evan-forbes commented 4 days ago

this is currently blocked by getting the trace data for an entire experiment from two nodes with latency.