Autonomic task scheduling algorithm for dynamic workloads through a load balancing technique for the cloud-computing environment

Applying the load balancing technique to allocate requests that dynamically enter the cloud environment is contributive in maintaining the system stability, reducing the response time, and increasing the resource productivity. One of the main challenges in dynamic load balancing is that it increases inter-VM communication overheads (swapping files between VMs). In most of the methods proposed for load balancing the issue of communication overheads is overlooked. Attempt is made here to address this problem through the Autonomous Load Balancing method. In the available studies on task scheduling in cloud computing, the focus is mostly on CPU-bound requests. Here, based on the resources, the needed the requests are divided into CPU-bound and I/O-bound requests. Considering both types of requests leads to the inability to apply the available load balancing methods. The CloudSim tool is applied here to evaluate this proposed method, which is then compared with Round Robin, Autonomous, Honey-Bee and Naïve Bayesian Load Balancing approaches. The results for the actual workloads of the NASA and Calgary servers and sample workload indicate that upon an increase in the requests and their variations together with heterogeneity of different VMs, this proposed algorithm can distribute the workload among them equally and allocate requests to appropriate VMs based on the required resources; thus, a decrease in the communication overheads and an increase in load balancing degree.