Date Added: Jul 2011
A load balancing mechanism for large-scale systems should be distributed and dynamic in order to accomplish scalability and high availability. Also, it should be autonomic in order to ease network management. The recent development in utility computing architectures, such as the so-called cloud computing platforms, has increased the demand for such mechanisms. This paper investigates a novel approach, based on the concept of virtual magnetic fields, by which ready-to-start tasks launched on a network middleware are "Attracted" to idle nodes. The key issue on such approach is that the update of workload information amongst the cooperating nodes of a network must be a low-cost and an autonomic operation.