Heuristic-based load balancing for identical virtual machines: a fair scheduling approach using probabilistic methods

The work that is considered in this article is the difficult task of designing a good scheduling policy for assigning the many activities involved in making multiple products to the available infrastructure of virtual machines (VMs). The objective of this study is to achieve an even workload entre virtual machines that are in charge of running manufacturing tasks. The ultimate goal of the research is to come up with a complete scheduling of all virtual machine-based tasks, with special attention to optimisation of a balanced level of lifetime as well as reduced variance between various virtual machines, which participate in the production activities. The main goal of this experiment is to mitigate the disparities in the turnaround time of VMs by allowing maintenance, or task handover to be done in a well-attuned way for a more uniform operational smoothness. It requires the optimization of the lowest work cycle of a virtual machine, which is very important to sustain the effective capabilities over time. Balancing the operational fairness of the VMs is recognised as the best scheduling policy for dealing with the problem difficulties. In this work, six different heuristic-based approaches are presented as feasible methods to solve the problem, based on mathematical formulations to provide a range of approximate solutions to the issues analyzed. The proposed approach is characterized by a probabilistic and iterative approach aimed at reinforcing the reliability of the obtained results. The results confirm that the proposed approximate solutions are effective, based on strict tests over 1,250 instances, with fixed metrics used to facilitate a comparison with all the heuristic algorithms. The experiments demonstrate that the repetitive-probabilistic heuristic dominates the other proposed heuristics in 82.2% of all instances, resulting in an average gap of 0.11 and time consumption of 0.036 s. The second-best heuristic, repetitive-mixed probabilistic heuristic, obtains 59.0% in percentage terms, the average gap is 0.28, and the running time is 0.034 s.