A Novel Hybrid Genetic Algorithm for Deadline Based Cost Optimization in Cloud Environment

Cloud computing is a well-established paradigm that is gaining popularity due to its high performance, elastic, and scalable computing resources over the internet. These features make it an it an appropriate platform to schedule complex scientific workflows over it. Workflow scheduling is considered NP-hard in nature. Scientists face a challenge when scheduling scientific workflow in a cloud due to its security, heterogeneity, and dynamic nature. Although there have been many studies on workflow scheduling in relation to grid and cluster computing throughout the years, not many of them have been specifically designed for the cloud. Furthermore, the earlier research did not take into account the fundamentals of cloud infrastructure, like varying performance of Virtual Machine (VM), acquire delay, virtual machine (VM) startup and shutdown times, bandwidth of channel, communication delay, and other QoS parameters. To fulfill the deadline and minimize the total computation cost, this research suggests a hybrid model based on Genetic Algorithms (GA) with appropriate heurist techniques. Heuristic techniques aid in producing high-quality seed to improve the performance of GA in its convergence rate. For this, we have applied Predict Earliest Finish Time (PEFT) and Heterogeneous Earliest Finish Time using Task Duplication (HEFT-TD) during the initialization of the population. We have used cutting-edge techniques in the genetic algorithm's crossover and mutation operators. Our hybrid paradigm considered varying performance of virtual machine, acquisition delay, booting time, and shutdown time of VMs. We performed our experiment on the simulator WorkflowSim on four types of scientific workflows. The outcomes show that the recommended approach performs workflow scheduling with the lowest costs and makespan when compared to other algorithms, demonstrating its exceptional superiority.