Reducing Latency in Tor Circuits with Unordered Delivery

Michael F. Nowlan, David Wolinsky, and Bryan Ford
Yale University

3rd USENIX Workshop on Free and Open Communications on the Internet (FOCI '13)
August 13, 2013, Washington, D.C.


Tor, the popular anonymous relay tool, incurs significant latency costs—partly due to extra network hops, but also due to TCP’s strict in-order delivery. We examine the problem of TCP’s head-of-line blocking in Tor, although this problem affects any application multiplexing streams atop TCP. Using uTCP and uTLS, techniques for enabling unordered delivery in TCP and TLS, respectively, we eliminate head-of-line blocking between Tor circuits sharing a TCP connection, without sacrificing Tor’s security. The small code footprint of uTCP and uTLS, and the minimal changes required to Tor, suggest the feasability of our approach. A micro-benchmark indicates that the integration of uTCP and uTLS can noticeably lower application-perceived latency.

Paper: PDF Slides: PowerPoint, PDF

This research was conducted with Government support under and awarded by DoD, Air Force Office of Scientific Research, and National Defense Science and Engineering Graduate (NDSEG) Fellowship, 32 CFR 168a. Additionally, this material is based upon work supported by the Defense Advanced Research Agency (DARPA) and SPAWAR Systems Center Pacific, Contract No. N66001-11-C-4018.