DMMP Seminar

Jasper de Jong — Decentralized Mechanisms for Throughput Scheduling

Motivated by the organization of decentralized service systems, we study new models for throughput scheduling. In throughput scheduling, we have a set of jobs i with value w_i, processing requirement p_i, and deadline d_i, to be processed non-preemptively on a set of unrelated servers. The goal is to maximize the total value of jobs finished before their deadline. While several approximation algorithms with different performance guarantees exist for this and related models, we are interested in decentralized mechanisms where the servers act selfishly according to some given, simple protocol. We show by simple, combinatorial arguments that, when each server deploys an α-approximation locally, any Nash equilibrium still yields an (α+1)-approximation with respect to the global optimum. This bound is tight, even in the case of related machines, unit weights and unit processing times. For models with identical machines, the bound can be improved to e^1/α/(e^1/α-1). Some of our results also extend to online models with corresponding competitive ratios.