Author
Listed:
- Sara Adnan Mahmood
- Hussein Najm Abd Ali
Abstract
Cloud task scheduling must balance conflicting objectives: deadline compliance, energy consumption, makespan, and resource utilization. A common design pairs a forecasting module and a reinforcement-learning placement agent with a periodic evolutionary refinement stage, usually a single-objective genetic algorithm that merges the four objectives into one weighted score. This paper studies whether replacing that stage with the reference-point-based multi-objective Non-dominated Sorting Genetic Algorithm III (NSGA-III) changes the scheduling results when forecasting and placement remain unchanged. Both strategies were evaluated on simulated five-machine clusters with workload-prediction signals from real monitoring traces over ten random seeds at 50,000 tasks (NSGA-III also at 100,000). The elitist genetic algorithm produced a mean of 0.495 percent of tasks violating their service-level agreement (SLA), against 0.178 percent for NSGA-III; a paired Wilcoxon signed-rank test confirmed the difference (p = 0.0039, r = 0.854). Doubling the workload left NSGA-III’s violation rate stable (paired p = 0.426). NSGA-III needed far more time per run (48.56 versus 9.05 minutes) and slightly more energy (30,582,411 versus 30,034,047 simulation units; p = 0.0020, r = 0.886). Sensitivity analyses on a 10,000-task protocol showed that the result depends on configuration: ten tasks per batch lay in a narrow working range, the number of reference points mattered little, and the final-selection weights mattered most. The two procedures also differ in search budget, but a genetic algorithm given NSGA-III’s budget and operators had a higher mean violation rate than either of them, so extra search does not explain the advantage. With task attributes derived from the Bitbrains traces, the ranking held, but the SLA advantage of NSGA-III shrank from 65 to 24 percent, and its energy penalty disappeared.
Suggested Citation
Handle:
RePEc:gvu:tcsst0:v:8:y:2026:i:4:id:2464
DOI: 10.36548/jtcsst.2026.4.005
Download full text from publisher
Corrections
All material on this site has been provided by the respective publishers and authors. You can help correct errors and omissions. When requesting a correction, please mention this item's handle: RePEc:gvu:tcsst0:v:8:y:2026:i:4:id:2464. See general information about how to correct material in RePEc.
If you have authored this item and are not yet registered with RePEc, we encourage you to do it here. This allows to link your profile to this item. It also allows you to accept potential citations to this item that we are uncertain about.
We have no bibliographic references for this item. You can help adding them by using this form .
If you know of missing items citing this one, you can help us creating those links by adding the relevant references in the same way as above, for each refering item. If you are a registered author of this item, you may also want to check the "citations" tab in your RePEc Author Service profile, as there may be some citations waiting for confirmation.
For technical questions regarding this item, or to correct its authors, title, abstract, bibliographic or download information, contact: Journal Manager (email available below). General contact details of provider: https://irojournals.com/tcsst .
Please note that corrections may take a couple of weeks to filter through
the various RePEc services.