Joint Scheduling of Pipeline Workflows and Real-Time Jobs with Varying Laxity Constraints

Abstract

The correct and timely execution of complex workloads in distributed systems relies on effective task scheduling mechanisms. This paper examines scheduling strategies for pipeline workflow applications coexisting with mixed-criticality periodic real-time tasks that exhibit heterogeneous laxity constraints. These real-time tasks can temporarily switch to imprecise execution modes, enabling pipeline jobs to progress with reduced preemption. Although higher-priority real-time tasks may interrupt PW execution, their criticality-dependent laxity provides additional scheduling flexibility that can be exploited without violating timing guarantees. Building on this observation, we propose two scheduling approaches that apply approximate execution to periodic real-time tasks while limiting the performance degradation of PW applications. Simulation-based experiments evaluate the impact of such approximations on pipeline workflow progress and overall system performance across a range of operating scenarios.