Conclusions and Future Work

• Architecture and algorithm (CSFQ) for high speed

  • achieve fair allocation close to FQ and comparable or better than FRED under most simulation scenarios
  • do not require core nodes to maintain per-flow state
  • can approximate weighted FQ

• Open problems

  • impact of very large latencies

• Future work

  • better estimators for flow and fair rates
  • per-flow guarantees without per flow management at core nodes

http://www.cs.cmu.edu/~istoica/csfq

Previous slide

Back to first slide

View graphic version