SMORE

a semi-oblivious traffic engineering system



Try it with YATES


Overview

SMORE is a semi-oblivious traffic engineering system which combines oblivious routing for static path selection with dynamic rate adaptation. SMORE uses oblivious routing to select a static set of paths which have:

  • low stretch for minimizing latency,
  • high diversity for ensuring robustness, and
  • good load balancing properties for achieving performance.
As demands change over time, SMORE dynamically load-balances traffic by re-adjusting the relative weights of these paths.

In traffic engineering, there is a fundamental trade-off between performance and robustness. Most systems are designed to optimize for one or the other, but few manage to achieve both. This challenge is further exacerbated by operational restrictions such as the number of paths, overhead due to churn, quantized splitting ratio imposed by hardware, etc.

SMORE achieves near-optimal performance in terms of congestion and load balancing metrics, is competitive with shortest-path based approaches in terms of latency, and is also robust, allowing traffic to be re-routed around failures without introducing bottlenecks while respecting operational constraints.

Talk (NSDI 2018)

Members

Cornell Cornell

CMU CMU

Facebook Facebook



USI USI Lugano

Publications

Praveen Kumar, Yang Yuan, Chris Yu, Nate Foster, Robert Kleinberg, Petr Lapukhov, Chiun Lin Lim, and Robert Soulé. Semi-Oblivious Traffic Engineering: The Road Not Taken. In USENIX Symposium on Networked Systems Design and Implementation (NSDI), Renton, WA, April 2018. [ conference version | slides ]

Praveen Kumar, Yang Yuan, Chris Yu, Nate Foster, Robert Kleinberg, Petr Lapukhov, Chiun Lin Lim, and Robert Soulé. Semi-Oblivious Traffic Engineering with SMORE. In ACM, IRTF & ISOC Applied Networking Research Workshop (ANRW), Montreal, Quebec, Canada, July 2018. [ extended abstract | slides ]

Praveen Kumar, Chris Yu, Yang Yuan, Nate Foster, Robert Kleinberg, and Robert Soulé. YATES: Rapid Prototyping for Traffic Engineering Systems. In ACM SIGCOMM Symposium on Software-Defined Networking Research (SOSR), Los Angeles, CA, March 2018. [ conference version | slides ]

Support

SMORE is partially supported by the National Science Foundation under Grant Number CCF-1637532 and by the Office of Naval Research under grant N00014-15-1-2177. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of these organizations.

  • NSF
  • ONR