Nidhi Chahal,
Purshotam Jha,
Sonia,
- Assistant Professor, Department of Electronics and Communication Engineering, Chandigarh Engineering College-CGC, Punjab, India
- Student, Department of Electronics and Communication Engineering, Chandigarh Engineering College-CGC, Punjab, India
- Student, Department of Electronics and Communication Engineering, Chandigarh Engineering College-CGC, Punjab, India
Abstract
Heterogeneous parallel computing systems are made up of a combination of multicore CPUs, GPUs, and specialized devices to boost computational power and utilize resources. But changes in workload intensity and processor capability, energy consumption and system conditions are dynamic, making efficient task scheduling difficult. Traditional scheduling approaches are based primarily on task properties or the state of the resources employed and can be sub-optimal when workloads are evolving. This paper introduces a predictive and feedback-based scheduling framework, named AI-PASS (AI-based Predictive, Adaptive and Sustainable Scheduler), which integrates workload prediction with adaptive taskresource selection. A novel Adaptive Task Suitability Score (ATSS) is used to predict execution time, processor utilization, energy usage, thermal condition, communication overhead and compatibility of tasks and resources. The weights associated are dynamically changed based on the condition of the system. AI-PASS can be described as being closed loop with monitoring, prediction, characterization of task, selection of resources, execution, measurement, and correction based feedback. The goal of the framework is to minimize makespan, energy usage and resource imbalance, and increase utilization. Further, simulation experiments are conducted to compare the performance of AI-PASS with Round Robin, Min-Min, and HEFT on heterogeneous CPU-GPU workloads which are measured in terms of makespan, energy consumption, throughput, CPU utilization, load imbalance and scheduling overhead.
Keywords: Heterogeneous computing, parallel computing, task scheduling, artificial intelligence, workload prediction, energy-aware scheduling, CPU-GPU scheduling, adaptive scheduling, resource allocation, ATSS
[This article belongs to Journal of Operating Systems Development & Trends ]
References
- Topcuoglu H, Hariri S, Wu MY. Performance-effective and low-complexity task scheduling for heterogeneous computing. IEEE Trans Parallel Distrib Syst. 2002;13(3):260-74.
- Liu Y, Qu H, Chen S, Feng X. Energy efficient task scheduling for heterogeneous multicore processors in edge computing. Sci Rep. 2025;15(1):11819.
- ul Huda Z. An enhanced hybrid PSO-LSTM framework for energy-aware task scheduling in heterogeneous cloud computing environments. J Emerg Paradigms Comput Syst. 2025;1(1).
- Sun C, Yang T, Lei Y. DDDQN-TS: a task scheduling and load balancing method based on optimized deep reinforcement learning in heterogeneous computing environment. Int J Intell Syst. 2022;37(11):9138-72.
- Shen W, Lin W, Wu W, Wu H, Li K. Reinforcement learning-based task scheduling for heterogeneous computing in end-edge-cloud environment. Cluster Comput. 2025;28(3):179.
- Nesmachnow S, Dorronsoro B, Pecero JE, Bouvry P. Energy-aware scheduling on multicore heterogeneous grid computing systems. J Grid Comput. 2013;11(4):653-80.
- Moulik S, Chaudhary R, Das Z. HEARS: a heterogeneous energy-aware real-time scheduler. Microprocess Microsyst. 2020;72:102939.
- Moulik S, Das Z, Devaraj R, Chakraborty S. SEAMERS: a semi-partitioned energy-aware scheduler for heterogeneous multicore real-time systems. J Syst Archit. 2021;114:101953.
- Lin Z, Li C, Tian L, Zhang B. A scheduling algorithm based on reinforcement learning for heterogeneous environments. Appl Soft Comput. 2022;130:109707.
- Chen J, He Y, Zhang Y, Han P, Du C. Energy-aware scheduling for dependent tasks in heterogeneous multiprocessor systems. J Syst Archit. 2022;129:102598.
- Chen G, Qi J, Sun Y, Hu X, Dong Z, Sun Y. A collaborative scheduling method for cloud computing heterogeneous workflows based on deep reinforcement learning. Future Gener Comput Syst. 2023;141:284-97.
- Song Y, Li C, Tian L, Song H. A reinforcement learning based job scheduling algorithm for heterogeneous computing environment. Comput Electr Eng. 2023;107:108653.
- Deng Z, Yan Z, Huang H, Shen H. Energy-aware task scheduling on heterogeneous computing systems with time constraint. IEEE Access. 2020;8:23936-50.
- Yao Z, Ding L, Zhang H, Li H, Chen L, Li Z. Deep reinforcement learning for job scheduling on load-aware heterogeneous cluster. J Big Data. 2025;12(1):276.
- Idouar Y, Cassagne A, Pilla LL, Sopena J, Bouyer M, Orhan D, et al. Energy-aware scheduling strategies for partially-replicable task chains on heterogeneous processors. Parallel Comput. 2026:103206.
- Tan Q, Chen W, Liu D. Task scheduling for heterogeneous multi-core processors based on deep reinforcement learning. Int J Intell Syst. 2025;2025(1):7562400.
- Idouar Y, Cassagne A, Pilla LL, Sopena J, Bouyer M, Orhan D, Lacassagne L, Galayko D, Barthou D, Jego C. Energy-aware scheduling strategies for partially-replicable task chains on heterogeneous processors. Parallel Computing. 2026 Jul 2:103206.
Journal of Operating Systems Development & Trends
| Volume | 13 | |
| Issue | 02 | |
| Received | 10/09/2026 | |
| Accepted | 15/09/2026 | |
| Published | 20/09/2026 | |
| Publication Time | 10 Days |