TPMCD: A method to optimizing cost and throughput for clustering tasks and hybrid containers in the cloud data center

Abstract

The regulatory of task classification or clustering and hybrid containers in cloud data centers has a lower overhead of cost compared to virtual machines, also it has a direct impact on the load balance, accessibility of virtual machines, and increase of efficiency. Therefore, additional resources with high computing power usage are one of the important issues. In the proposed method merging the index parameters of response time, execution accuracy and their sensitivity rate have been used. In TPMCD(ThroughPut and cost optimizing Method for Clustering tasks and hybrid containers in the cloud Data center), customers agreement, as a service and performance of the connection, the efficiency of service quality and reliability of algorithms, requests, and confirmations (short, medium, long) due to the configuration of resources and containers and the intelligent detector threshold, protection of the increase in system efficiency and energy consumption decrease synchronously against dynamic workloads and changes in user requests. Classification and re-clustering of tasks in the algorithm have led to an improvement in the real execution time compared to the execution time of the studied algorithms. In the proposed method, by correctly allocating resources for scoring unbalanced data for allocating resources and applications and communicating between containers. In TPMCD, parameters of weight, size, and scoring are used in assigning tasks to processing resources. Confidence interval has been done in proposed method due to the possibility of a small difference in scheduling between different virtual machines. In the TPMCD algorithm, choosing the right VM and reducing the critical points, in the hosts where the load imbalance is created, the load balance is optimized by considering the sensitivity rate and scoring the average tasks. TPMCD method have optimized time and cost by decreasing redundancy. From the obtained results in the evaluation, this method performed better than other ones 7% in cost, 4% in throughput, and 9.5% in real execution time on average simultaneously. Finally, the proposed approach was 3% better than the KC method in the number of nodes used.