Quantifying the Carbon Reduction of DAG Workloads: A Job Shop Scheduling Perspective

2025-12-08

AI Summary

Existing carbon-aware schedulers treat jobs as atomic units, ignoring their DAG structure and heterogeneous resource demands across subtasks, thereby limiting carbon efficiency. Method: We model batched DAG workloads from a job-shop scheduling perspective and propose a dependency-aware carbon-aware scheduling framework that maximizes execution of critical-path tasks during low-carbon time intervals—without extending the optimal makespan. Leveraging a flexible job-shop formulation and an offline solver, we quantify the theoretical carbon-reduction upper bound attainable through structured scheduling. Contribution/Results: Our approach achieves an average 25% reduction in carbon emissions; permitting twice the optimal makespan nearly doubles emission reductions. We explicitly characterize the Pareto trade-offs among carbon emissions, energy consumption, and makespan. Experiments demonstrate that explicit modeling of task structure and server scale are decisive factors for carbon efficiency—establishing, for the first time, the fundamental limits and design principles of carbon-aware DAG scheduling.

Technology Category

• Planning, Routing, and Scheduling: Scheduling under Uncertainty
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Machine Learning: Hardware-aware ML

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Sustainability and carbon-aware systems for Web, mobile, and WoT
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems

Abstract

Carbon-aware schedulers aim to reduce the operational carbon footprint of data centers by running flexible workloads during periods of low carbon intensity. Most schedulers treat workloads as single monolithic tasks, ignoring that many jobs, like video encoding or offline inference, consist of smaller tasks with specific dependencies and resource needs; however, knowledge of this structure enables opportunities for greater carbon efficiency. We quantify the maximum benefit of a dependency-aware approach for batch workloads. We model the problem as a flexible job-shop scheduling variant and use an offline solver to compute upper bounds on carbon and energy savings. Results show up to $25%$ lower carbon emissions on average without increasing the optimal makespan (total job completion time) compared to a makespan-only baseline. Although in heterogeneous server setup, these schedules may use more energy than energy-optimal ones. Our results also show that allowing twice the optimal makespan nearly doubles the carbon savings, underscoring the tension between carbon, energy, and makespan. We also highlight key factors such as job structure and server count influence the achievable carbon reductions.

Problem

• Modeling batch workloads as flexible job-shop scheduling to quantify carbon reduction potential
• Evaluating dependency-aware scheduling for lower carbon emissions without increasing makespan
• Analyzing trade-offs between carbon savings, energy use, and job completion time

Innovation

• Modeling DAG workloads as flexible job-shop scheduling
• Using offline solver to compute carbon savings upper bounds
• Allowing makespan increase to double carbon reductions