Research Article
Computational Mathematical Model for Resource Scheduling in Cloud Computing Environments: Integrating Integer Programming and Reinforcement Learning
@ARTICLE{10.4108/eetsis.11134, author={Zhiyan Li and Xiaohui Zhang and Baoxia Jin}, title={Computational Mathematical Model for Resource Scheduling in Cloud Computing Environments: Integrating Integer Programming and Reinforcement Learning}, journal={EAI Endorsed Transactions on Scalable Information Systems}, volume={12}, number={8}, publisher={EAI}, journal_a={SIS}, year={2026}, month={3}, keywords={Cloud Computing, Resource Scheduling, Integer Programming, Reinforcement Learning, Predictive Optimization}, doi={10.4108/eetsis.11134} }- Zhiyan Li
Xiaohui Zhang
Baoxia Jin
Year: 2026
Computational Mathematical Model for Resource Scheduling in Cloud Computing Environments: Integrating Integer Programming and Reinforcement Learning
SIS
EAI
DOI: 10.4108/eetsis.11134
Abstract
Cloud computing, on the other hand, is very fast and very easy to scale while still having to rely on traditional scheduling methods that cause delays, SLA violations, and a lack of flexibility, mainly when the workloads change. These problems occur in large-scale environments that are very inefficient and hence consume a lot of resources and power. A hybrid solution combining Integer Programming (IP) and Reinforcement Learning (RL) is offered by the authors to make the resource allocation adaptive, energy-efficient, and at the same time provide SLA compliance. The system is supplied with a real Cloud Workload Dataset that offers comprehensive details at both task and system levels. The first step involves Integer Programming, which is used to produce a constraint-aware basic allocation and to ascertain its feasibility by considering not only resource availability but also the limitations imposed by deadlines. At this point, RL takes over from the allocation and thereby improves it further by utilizing Q-learning and ε-greedy policy for making real-time adjustments according to the states of the system and feedback. The interdependence between learning and optimization is thus continuous, leading to better scheduling outcomes over time. The system's performance is evaluated through the application of standard cloud performance metrics such as SLA compliance, resource utilization, task completion time, energy efficiency, and makespan. The hybrid IP–RL framework achieves 98.6% SLA compliance, 99.6% allocation efficiency, and a 12–18% reduction in task completion time when compared to the baseline. Furthermore, it requires 15% less energy and achieves a 92.4% better makespan efficiency. The given results prove that the hybrid method beats the traditional scheduling models in performance concerning all the metrics and opens a door for a cloud resource scheduling that is both scalable and adaptive.
Copyright © 2026 Zhiyan Li et al., licensed to EAI. This is an open access article distributed under the terms of the CC BY-NC-SA 4.0, which permits copying, redistributing, remixing, transformation, and building upon the material in any medium so long as the original work is properly cited.
