Abstract :

Huge amounts of data are proliferating through communication networks resulting in congestion of nodes on the networks, which varies according to peak usage times during the day. This work proposes a solution aimed at alleviating the burdened nodes, thereby improving the flow of traffic throughout the networks hence improving the users' usage experience and thereby productivity and efficiency of carrying out their work assignments. Given the explosive growth in network traffic, applying swarm optimization algorithms appears more appropriate than other techniques such as Round Robin and Weighted Least Connection to optimize congestion problem from source node to destination node. The idea is to mitigate the traffic load burden during peak hours using meta-heuristic approach about network optimization in IoT. Due to increasing massive amounts of data traffic on computer network the decentralized approach is attracting the researchers' attention. The swarm based algorithms make use of multimodal functions to find the network peaks. The higher the peaks, the greater the chance of traffic on the respective edges, which is calculated by the objective function of swarm intelligence algorithms. Another component of swarm based algorithms is the fitness measure, such as the concentration of pheromone in ACO, and luciferin level in GSO, that controls the queue of tasks accumulated on particular nodes. The main objective of load balancing method is to allocate the traffic load uniformly dispersed to achieve best performance on homogeneous and heterogeneous data networks. Load balancing is used to enhance user fulfillment, better utilization of resource, execution time and waiting time of task coming from several locations/nodes. The load balancer needs able to predict the load on those edges having lesser amounts of traffic. The proposed solution aims to alleviate the burdened nodes, thereby improving the flow of traffic throughout the networks hence improving the users' usage experience and thereby productivity and efficiency of carrying out their work assignments. This work studies three swarm intelligence algorithms, namely Particle Swarm Optimisation (PSO), Cuckoo Search (CK) and Glow-worm Swarm Optimisation (GSO) to the network load balancing problem. The burden of network traffic during peak traffic times is studied. The need is to find the shortest route from source to destination, identify the nodes having higher traffic density and divert the traffic randomly towards the lighter loaded nodes. The goal is to balance the queue length, memory load and the number of active connections to establish that the network is approximately balanced. The purpose is to maintain the performance factor. In this regard the throughput is to be maximized as well as to minimize the response time, time out, and delay time to achieve stability on different nodes of the computer network. The GSO algorithm works in a decentralized manner by decreasing response time and network overhead in addition by increasing throughput to improve the health of connected virtual machines of neighboring nodes. The results produced by Glowworm Swarm Optimization (GSO) shows improvement of 71.17%, 74.14% and 84.15% under 50, 100 and 200 nodes in peak hour. In comparison with Particle Swarm Optimization (PSO), GSO shows improvement by 13.87%, 11.75% and 23.72%, and 10.61%, 3.19% and 6.00% with Cuckoo Search (CK) algorithm for networks with 50, 100 and 200 nodes respectively.