New Directions in Stochastic Scheduling

  • by

Mor Harchol-Balter, Carnegie Mellon University

The massive expansion of datacenter computing has led to a plethora of new queueing models and new scheduling problems.  This talk will discuss several new open problems of practical importance.  Topics include:

  1. Advancing the state of scheduling for the M/G/1
  2. Scheduling for multiserver (M/G/k) systems
  3. Queueing for today’s multiserver jobs
  4. Scheduling of malleable jobs, with flexible parallelizability.

In addition we will characterize today’s datacenter workloads, particularly their extremely heavy tails.

This talk is based on the following paper: “Open problems in queueing theory inspired by datacenter computing.”  Queueing Systems, vol. 97, no. 1, 2021, pp. 3-37.

The presentation slides are availible below: