A meta-heuristic approach-aided multi-objective strategy with optimal resource allocation via fault tolerant and priority-based scheduling for load balancing in cloud

Cloud computing has been referred to as a successive high-functionality computing sector in the past years for organizations and also individuals. Numerous major previous experiments on the cloud computing sector have existed but there are still specific problems with the distribution of workload in the system of cloud networks. Effective task allocation is complex due to the existence of various Virtual Machines (VMs) and resources. In these approaches, the CSPs must ensure better service delivery functionality, preventing complexities like under or overloaded hosts. These may increase the processing period or result in machine damage. The primary concept of the recommended mechanism is to develop load balancing in the cloud computing sector with optimal resource allocation. With the quick network utilization and also the traffic of data in the cloud sector, the intensity of the computation and the power of processing are improved. In order to decrease the pressure of the network and to enrich the efficacy of the computation the load balancing mechanism is significant. In the starting phase, the scheduling process is executed with the utilization of fault tolerance and a priority-based scheduling process. Then, the optimization of the resources in the scheduling is done with the Modified Random Parameter Dragonfly Algorithm. To further evaluate the load balancing model, the objective function is derived by makespan, average resource utilization, throughput, delay, success rate, and execution time. On the other hand, owing to the dynamic nature of the model, the server (resource) status is varied continuously. Thus, in order to assign the task to the VM resources or server and to efficiently balance the load in the cloud model, the current status of the server is predicted by the Attention-based Bilateral Long Short-term Term Memory (ABi-LSTM) model before resource allocation. From the result analysis, the throughput of the developed model is 95.5, and also the throughput value of the existing methods such as JAYA, SCO, MFOA, and DA is 91, 89, 92, and 90.5. Finally, this model schedules the resources for effective scheduling based on the resource or load demand and priority of the tasks. In the end, the performance is evaluated using various performance metrics.