Load Balancing of Mobile Agents Based Applications in Grid Systems

Advanced grid functionalities and applications can be developed exploiting mobile agents technology. State-full migration can help to distribute computation where data reside and among new available and less busy machines. However when mobile agent technology is used new kinds of interaction facilities, such as cloning and migration, require execution of agent platforms, which represent an additional software layer that affects system performances. Also underlying hardware architecture can affect agent platform behaviour in an unexpected way. It means that is necessary to predict dynamics of workload distribution and to monitor system utilization in order to design effective load balancing strategies. We propose here a software architecture that, using results of system monitoring, applies a load balancing strategy to maintain in a fixed range threshold performances of a mobile agents based application.