Mean response time analysis of the Randomized Timers policy

Establish a rigorous mean response time analysis for the Randomized Timers scheduling policy within the Multiserver Job (MSJ) model defined in this paper, where class‑i jobs arrive according to independent Poisson processes and have i.i.d. exponential service durations. Quantify the mean response time under Randomized Timers for multiserver jobs with heterogeneous server needs, and characterize its performance relative to stability conditions.

Background

The paper surveys non-preemptive scheduling policies for multiserver jobs and highlights Randomized Timers as a throughput-optimal, non-preemptive policy derived from MaxWeight. Despite its stability guarantees, the literature lacks a mean response time analysis for Randomized Timers, limiting its practical evaluation and comparison against other policies.

This work develops Most Servers First with Quickswap (MSFQ) and provides stability and mean response time analysis for MSFQ in the one-or-all case, emphasizing the need for similar response-time guarantees for alternative non-preemptive, throughput-optimal policies such as Randomized Timers.