Multi-agent Based Network Performance Tuning in Grid

Grid environment is used for performing high-end computations with a large number of systems. The proposed work is to monitor the grid network with the help of mobile agents and tune the network metrics on performance degradation of the grid network. The grid network is monitored using the cost function to analyze the network performance in grid. Any degradation in the network performance is reflected by the cost function. The network parameter that has caused the degradation will then be identified through analyzing of each network metrics by the tuning manager. Then the network performance is tuned by tuning the degraded network metrics, which has high impact over it. Tuning is done based on changing the sending and receiving TCP socket buffer size of the compute nodes. It is also shown that the network monitoring is equally important as the resource metrics at the time of job submission in the grid network.