A very important issue in distributed systems is the management of load among system nodes. Load balancing enables a better use of the system resource capabilities and a better performance by allocating nodes that are more suitable for the execution of some tasks. This work proposes a multi-agent load balancing architecture for distributed object applications called MALBA which uses artificial neural networks and a set of policies to support the process of migrating and replicating the application objects. MALBA architecture outlines a decentralized balancing process that balances the load among nodes which execute a distributed object application.

No abstract