Migrate when necessary: toward partitioned reclaiming for soft real-time tasks

TL;DR

This paper introduces a partitioned reclaiming strategy for soft real-time tasks on multi-core systems, allowing temporary task migration and load balancing to reduce missed deadlines, validated through simulations.

Contribution

It proposes a novel reclaiming approach with task migration and load balancing for soft real-time multi-core scheduling, improving deadline adherence over existing methods.

Findings

Effective reduction in missed deadlines compared to other approaches

Load balancing improves task scheduling efficiency

Performance varies with different partitioning heuristics

Abstract

This paper presents a new strategy for scheduling soft real-time tasks on multiple identical cores. The proposed approach is based on partitioned CPU reservations and it uses a reclaiming mechanism to reduce the number of missed deadlines. We introduce the possibility for a task to temporarily migrate to another, less charged, CPU when it has exhausted the reserved bandwidth on its allocated CPU. In addition, we propose a simple load balancing method to decrease the number of deadlines missed by the tasks. The proposed algorithm has been evaluated through simulations, showing its effectiveness (compared to other multi-core reclaiming approaches) and comparing the performance of different partitioning heuristics (Best Fit, Worst Fit and First Fit).