A Novel Approach for Load Balancing in Distributed System using FIFO-Support Vector Machine (FIFOSVM)

Standard balancing algorithms used for load balancing in the distributed environment do not give best solution always. Latency in the system deter its performance, so it is important to optimize the load balancing strategies to get better performance. An optimum algorithm should distribute the load to the available servers based on their processing power. There are many load balancing algorithms available. Most load balancer dynamically chooses next eligible server. Simple load balancers use FIFO and round robin (RR) algorithms. Each algorithm would be efficient in one aspect and might be inefficient otherwise. In our proposed approach, we will use few algorithms such as FIFO with Support Vector Machine (SVM) and invoke them during a particular situation when they are efficient. Our simulation results show that our load balancer significantly improves the average and total response time of client tasks and thus increases the performance of the overall system.