Minimizing mean response time for multiserver jobs with fixed server requirements

Determine an optimal scheduling policy that minimizes mean response time for multiserver jobs—jobs that require a fixed number of servers—under general conditions, including in heavy-traffic regimes.

Background

Multiserver jobs are a related model where each job has a hard, fixed server requirement. The authors highlight that the mean response time minimization problem is not known in general even in heavy traffic.

This contrasts with the malleable job model studied in the paper, where jobs adapt their level of parallelism and stability concerns differ.