Improving resource management with multi-instance broker scheduling algorithm in hierarchical grid computing

This research addresses the performance issue of grid computing based on resource management, resource broker, resource scheduling and algorithm for middleware in the grid environment. The improved Hierarchical Load Balancing Algorithm (iHLBA) was chosen as the benchmark algorithm as it focuses on scheduling for hierarchical grid computing environment. The main purpose of this research is to propose a design of scheduling algorithm for hierarchical grid computing. For this purpose, processing performance, namely makespan time, response time, throughput, job distribution and the threshold, is the focus parameters of this study. Thus, Discrete Event Simulation is used as the research methodology in this study. In brief, this research aims to solve resource management issues so as to enhance the grid computing performance. The iHLBA scheduling algorithm was tested and it was found that the main issue was the delay occurring on the job submission. This issue needs to be addressed properly in order to obtain the targeted performance. Secondly, the receiving process of iHLBA is followed through on pre-defined settings without making any adjustment to the resource items of computing power on the destination node. Therefore, the issue that occurs is the grid performance in terms of throughput. Hence, resource items such as the CPU, memory and network bandwidth capacity have to be considered when gauging the grid computing power. The distribution of job into grid resources is an interesting aspect to be explored so as to gain highest fairness rate compared to the current performance. Multi-Instance Broker Scheduling Algorithm (MiBSA) has been proposed as a new scheduling algorithm to get rid of the drawback from the iHLBA algorithm. The Self-Adaptive Broker Manager, Multi-Instance Broker, Dual Layer Computing Allocation and Performance Recommender are the new modules introduced to achieve the research objectives. The designed modules is in the form of algorithms, GridSim, Java Development Kits and Java Language used to support the design process whereas the Netbeans software serves as the platform to simulate the experiment so as to obtain the results. Results from each of the variables and algorithms had been compared and statistical methods such as the t-test and normality test were used to verify the results. Overall, the final results showed that the new designed algorithm (i.e., MiBSA algorithm) has successfully improved grid processing performance and successfully surpassed the performance of previous algorithm