The VCG Mechanism for Bayesian Scheduling

Abstract: We study the problem of scheduling $m$ tasks to $n$ selfish, unrelated machines in order to minimize the makespan, where the execution times are independent random variables, identical across machines. We show that the VCG mechanism, which myopically allocates each task to its best machine, achieves an approximation ratio of $O\left(\frac{\ln n}{\ln \ln n}\right)$. This improves significantly on the previously best known bound of $O\left(\frac{m}{n}\right)$ for prior-independent mechanisms, given by Chawla et al. [STOC'13] under the additional assumption of Monotone Hazard Rate (MHR) distributions. Although we demonstrate that this is in general tight, if we do maintain the MHR assumption, then we get improved, (small) constant bounds for $m\geq n\ln n$ i.i.d. tasks, while we also identify a sufficient condition on the distribution that yields a constant approximation ratio regardless of the number of tasks.