frontmatter 
title: "Studying the Effect of Schedule Preemption on Dynamic Task Graph Scheduling"
venue: MILCOM 2025 - IEEE Military Communications Conference
doi: 10.1109/MILCOM64451.2025.11310446
date: 2025-10-06
authors:
  - Mohammadali Khodabandehlou
  - "[[Jared Coleman]]"
  - Niranjan Suri
  - Bhaskar Krishnamachari
links:
  - https://arxiv.org/pdf/2602.03081
abstract: "Dynamic scheduling of task graphs is often addressed without revisiting prior task allocations, with a primary focus on minimizing makespan. We study controlled schedule preemption, introducing the Last-K Preemption model, which selectively reschedules recent task graphs while preserving earlier allocations. Using synthetic, RIoTBench, WFCommons, and adversarial workloads, we compare preemptive, non-preemptive, and partial-preemptive strategies across makespan, fairness, utilization, and runtime. Results show moderate preemption can match most makespan and utilization gains of full preemption while maintaining fairness and low overhead."
tags:
  - distributed-computing
  - internet-of-things

Studying the Effect of Schedule Preemption on Dynamic Task Graph Scheduling

Mohammadali Khodabandehlou, Jared Coleman, Niranjan Suri, Bhaskar Krishnamachari

MILCOM 2025 - IEEE Military Communications Conference · October 6, 2025 · doi:10.1109/MILCOM64451.2025.11310446 · arXiv

Abstract

Dynamic scheduling of task graphs is often addressed without revisiting prior task allocations, with a primary focus on minimizing makespan. We study controlled schedule preemption, introducing the Last-K Preemption model, which selectively reschedules recent task graphs while preserving earlier allocations. Using synthetic, RIoTBench, WFCommons, and adversarial workloads, we compare preemptive, non-preemptive, and partial-preemptive strategies across makespan, fairness, utilization, and runtime. Results show moderate preemption can match most makespan and utilization gains of full preemption while maintaining fairness and low overhead.

Updated