CPU Scavenging Through the Swarm Intelligence of Autonomous Agents

"Clara" was the code name for this work. The word itself is a scientific name for a kind of ant species. In this work, we present a framework to support parallel computations, such as the one performed by the Seti@Home project, in which the workload may be subdivided in a large number of independent tasks. Unlike Seti@Home, however, Clara is not based on the master-slave paradigm, in which a well-known centralized master is responsible for supplying slave machines with computing tasks. In Clara, every node of the network is enabled to produce new tasks and introduce them in the network for computation. A swarm of Clara agents (ANTs) are responsible for exploring the network and balancing the workload by dispersing the tasks among all reachable nodes. Once computed, results are sent back to the originator node that may use appropriate mechanisms based on lease techniques to keep track of task assignments. Peculiarity of Clara comes from its ability of self-organization, adaptation and resilience which is not present in today s distributed computing paradigm.