DEVELOPING WISELY RANDOMIZED WEIGHTED THROTTLED LOAD-BALANCING ALGORITHM FOR CLOUD-COMPUTING ENVIRONMENT

Abstract

From other challenges of cloud computing, Load balancing is one of the main issues in the environment. Due to current vast demand on the technology, the load-balancing algorithms must have the ability to control the efficiency in distributing the loads across the resources. Not only this, but they must also ensure the usage if the resources in smart strategy so that some resources will not be on over-utilization state and others are in under-utilization. Many researchers proposed various load balancing and job scheduling algorithms in cloud computing, but there is still some inefficiency in the system performance and load imbalance.
The objective of this research is to develop efficient and fair load balancing algorithm for a cloud computing environment and compare the result of the proposed algorithm with the well-known load balancing algorithms. The efficiency of the proposed algorithm is based on response time and data processing time, and the fairness is based on resource utilization parameter.
Therefore, in this research, we propose a load balancing algorithm to improve the performance, efficiency, and distribution of load on a homogeneous cloud computing environment, in terms of response time and resource utilization. We propose an improved algorithm based on randomization and weight (by checking bandwidth). The algorithm has a wise random picking strategy during selecting two random Virtual Machines. This helps to have two unique Virtual Machines and to escape unnecessary checking of similar VMs’ bandwidth.
The efficiency of response time, resource utilization, and data processing time of the proposed algorithm is analyzed using CloudAnalyst simulator and compare with the other well-known and related existing algorithm called Randomized, Throttled, Randomized Weighted Throttled algorithms. The results showed improvements on average response time and resource utilization especially when we put much pressure by sending large amount of request-per-user.