Thanks for the nice presentation. I have got a question to your motivation:
– You pose the question “Can we do better than existing solutions”. My question is “Better in terms of what?” It is known that proportional fair scheduling operates on the Pareto-boundary of the system, i.e., one cannot at the same time achieve higher throughput and better fairness; one can only sacrifice one for the other. Please provide some comments to your view on this.
Thanks for your question, indeed it is very fair to think that way. There is always trade of between throughput and fairness. Proportional Fair Sharing is a maximization problem. We have proven in a full paper that our algorithm converges to the solution of an ordinary differential equation. ODE has a unique equilibrium. The existence of a unique equilibrium of ODE determine the throughput of each user and hence delay (Please also check “Convergence of Proportional-Fair Sharing Algorithms Under General Conditions” by Kushner). Since our algorithm convergence faster it achieves better throughput and fairness than PFS. Hope that answers your question.
Viewing 2 posts - 1 through 2 (of 2 total)
You must be logged in to reply to this topic.