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 set of paths which have:

  • low stretch for minimizing latency,
  • high diversity for ensuring robustness, and
  • good load balancing properties for achieving performance.

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.

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, 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 ]