Surprising Results on Task Assignment in Server Farms with High-Variability Workloads

This paper investigates the performance of task assignment policies for server farms, as the variability of job sizes (service demands) approaches infinity. Our results reveal that some common wisdoms regarding task assignment are flawed. The Size-Interval-Task-Assignment policy (SITA), which assigns each server a unique size range, was heretofore thought of by some as the panacea for dealing with high-variability job-size distributions. We show SITA to be inferior to the much simpler greedy policy, Least-Work-Left (LWL), for certain common job-size distributions, including many modal, hyperexponential, and Pareto distributions. We also define regimes where SITA's performance is superior, and prove simple closed-form bounds on its performance for the above-mentioned distributions.