An Ant Colony-based framework for Internet-scale Peer-to-Peer Grids
Francesco Palmieri · 2007
Grid is an exciting buzzword in the computing world today, mainly in the scientific area. It is usually defined as the exploitation of a varied set of networked computing resources, including large or small computers, PDAs, file servers and graphics devices. The emerging Internet based peer-to-peer Grid infrastructures, which are based on a “flat ” organization allowing seamless discovery, access to, and interactions among resources and services, have complex and highly dynamic computational and interaction behaviors, and when combined with the uncertainty of the underlying Internet infrastructure, result in significant development and management challenges. Accordingly, in this paper, we propose a multi-agent based Peer-to-Peer Grid computing architecture based on swarm intelligence and precisely on the ant-colony meta-heuristic, to map the solution capability of social insects to the peer node interaction and resource scheduling problem. The main characteristics of this architecture are highlighted by its promising performance and scalability, and its adaptive resource management and scheduling mechanisms. It is completely decentralized, allowing every node in the system to act as a client and a runtime server, by generating new jobs, submitting them to the network and participating, by running several jobs, to the overall computational resource sharing. A Grid infrastructure designed in this way may be interesting for individual peer entities that want to share their resources through the Internet within a large community in order to exploit the resulting computing power effectively. Neither class of current Grids is capable of scaling up to a so large user base, with a wide variety of applications and a plethora of user profiles. This will force distributed application designers to look for alternative, more suitable computational models and architectures.