Reorganization and discovery of grid information with epidemic tuning

摘要

This paper examines a multi-agent approach to spatially sort and discovery information about the resources offered by a Grid. Agents, whose behavior is inspired by ant colonies, replicate and distribute resource descriptors according to the class to which the corresponding resources belong. This facilitates resource discovery operations: query messages are attracted towards hosts that store information about a large number of resources having the required characteristics. The presented reorganization and discovery protocols feature self-organization and decentralization characteristics, since operations are performed only on the basis of local information. Agents can either replicate or simply relocate resource descriptors. These two operation modes are aimed, respectively, at fostering the dissemination or the reorganization of information. The balance between these two objectives can be modulated by setting the parameter of an ant-inspired pheromone mechanism. Balance can be static, i. e., decided a priori, or dynamic, in the case that user and network requirements change with time. In the latter case, an "epidemic" mechanism is used to communicate the value of this parameter to the hosts and agents of the Grid. Simulation analysis confirms the effectiveness of the reorganization and discovery protocols and of the mentioned epidemic tuning mechanism.