Enabling Long-term Fairness in Dynamic Resource Allocation

We study the fairness of dynamic resource allocation problem under α-fairness criterion. recognize two different objectives that naturally arise in this problem: well-understood slot-fairness objective aims to ensure at every timeslot, and less explored horizon-fairness across utilities accumulated over a time horizon. argue comes lower price terms social welfare. with regret as performance metric show vanishing cannot be achieved presence an unrestricted adversary. propose restrictions on adversary's capabilities corresponding realistic scenarios online policy indeed guarantees these restrictions. demonstrate applicability proposed framework representative management considering virtualized caching system where caches cooperate serve content requests.