Comprehensive Learning Particle Swarm Optimization(CLPSO) based hybrid cloud task scheduling with predefined completion time constraints

This paper proposed comprehensive learning PSO based on deadline constraint task scheduling for a hybrid IaaS cloud environment. One of the significant challenges in managing cloud services is scheduling tasks to VMs in such a way as to optimize the objective (that may be a reduction in make span, idle time, and waiting time or increasing the profit, availability, and serviceability). In this study, profit maximization is taken as an objective. This study concentrates on the workload on a batch, i.e., a bag of independent tasks which can be a kind of image rendering process, simulation, data processing, or video rendering process. We have used CLPSO for task scheduling and fixed the fitness function as profit maximization. The particle is designed so that we can easily obtain the allocation sequence by ranking the particle's coordinates. We found that for the given set of VMs (private and external cloud), the optimal or best number of tasks that can be taken is 3 to 5, as the average profit for those tasks is higher than the other set of tasks. Similarly, for a different set of cases, we can run the same setup and find the best number of tasks that can be accommodated to increase the profit. This study can be extended by changing the objective and can be repeated to get the inference on the new objective.