wiki:org/meetings/2015WinterDevMeeting/Notes/TorPerformance

Tor Performance

Datagram Transport

  • Flow control
    • Want non-interference but still congestion control
  • Survey Paper from Ian's student Michelle
    • "Soon" - ACM CSUR

Peerflow

  • Vulnerable to an inflation attack
    • Feedback problem if you control > ~3% traffic
    • Unknown solution as of yet. May be detectable

TorFlow hacks/issues

  • Could only measure middle nodes
  • Could vary exit destination
  • Could use dirport or echo protocol hacks
  • Concurrency improvements, but may cause interference?
    • Would use OONI/txturn
  • Consensus updates are slow
  • Unmeasured relays may get grouped together at 1.0 ratio.
    • Mix in measured ones

TorRTT

  • Stream-based RTT
    • 1 vs 3 vs 5 measurements
Last modified 4 years ago Last modified on Mar 5, 2015, 5:11:40 PM