Abstract

Cloud computing relies heavily on load balancing to distribute workloads evenly among servers, network connections, and drives. The cloud system has been assigned some load which can be underloaded, overloaded, or balanced depending on the cloud architecture and user requests. An important component of task scheduling in clouds is the load balancing of workloads that may be dependent or independent of virtual machines (VMs). To overcome these drawbacks, a novel Load Balancing of Virtual Machine (LBVM) in Cloud Computing has been proposed in this paper. The input tasks from multiple users were collected in a single task collector and sent towards the load balancer, which contains the deep learning network called the Bi-LSTM technique. When the load is unbalanced, the VM migration will begin by sending the task details to the load balancer. The Bi-LSTM is optimized by a Genetic Expression Programming (GEP) optimizer and finally, it balances the input loads in VMs. The efficiency of the proposed LBVM has been determined using the existing techniques such as MVM, PLBVM, and VMIS in terms of evaluation metrics such as configuration latency, detection rate, accuracy etc. Experimental results shows that the proposed method reduces the Migration Time of 49%, 41.7%, and 17.8% than MVM, PLBVM, VMIS existing techniques respectively.

Introduction

Cloud services that may be accessible online are how computational hardware and software resources are packaged in cloud computing. There are three different types of cloud computing applications that have been developed: infrastructure as a service (IaaS), platform as a service (PaaS), and software as a service (SaaS) [ [1] , [2] , [3] ]. In order to optimize [ [4] , [5] , [6] ] streamline operations and deliver acceptable levels of speed for the consumers, handling massive data sets necessitates a number of strategies. By maintaining effective management of cloud resources, it is possible to accomplish the efficient and scalable properties of cloud computing [ 7 ]. Users might cite better utilisation of distributed resources and their application to increase throughput [ 8 ], performance [ 9 ], and problem-solving at a big scale [ 10 ] as the primary objectives of cloud computing (see Fig. 2 , Fig. 3 , Fig. 4 , Fig. 5 , Fig. 6 , Fig. 7 ).

The chances of failures that could simultaneously harm the services in cloud systems are reduced by the use of load balancing [ 11 ] and redundant mirrored databases in cluster techniques [ 12 ], which span several availability zones. The load balancer can move to another resource if one system has an outage [ 13 ]. By maximising resource availability and minimising the amount of downtime experienced by organisations during outages, load balancing techniques among the circumstances towards cloud computing help to lower expenses related to document management systems [ 14 , 15 ].

Conclusion

Load Balancing of Virtual Machine (LBVM) in Cloud Computing has been proposed in this paper. The input tasks from multiple users were collected in a single task collector and sent towards to the load balancer, there it perform the Bi-LSTM technique by using Genetic Expression Programming (GEP) optimization algorithm and the collected schedules sent to Virtual Machines (VMs) and finally it balances the input loads in VMs, When the load was unbalanced, VM migration was done. According to the results, better prediction is provided by Bi-LSTM based modelling than by traditional LSTM based models because it is based on extra data training, provides better prediction than conventional LSTM based models. In load balancing, the proposed Bi-LSTM and GEP optimization are performed, the Bi-LSTM is the optimized by GEP optimizer. The optimized cloud resources are goes to VMs which performs in cloud based and there the VMs balances the tasks. The proposed method has been evaluated in terms of MVM, PLBVM, VMIS and the proposed Load Balancing of Virtual Machine (LBVM). The proposed method reduces the Migration Time of 49%, 41.7%, 17.8% than MVM, PLBVM, VMIS existing techniques. Because cloud computing offers virtualization, scalability, infinite computational resources, and the capacity to store massive volumes of both organized and unstructured data, it is perfect for deep learning. We intend to enhance this type of load balancing for buildings that have dependent activities in the future.