A Scalability Analysis of Grid Allocation Mechanisms

Grid Computing is a paradigmatic application forthe requirements associated with the exponentially growingcomplexity in the engineering, operation and maintenance oftoday’s information systems. Grid Computing comprises evergrowingglobal communication infrastructures and millions ofpossible system elements into the picture. Self-organizing, agentbasedGrids share many properties and research questionswith Autonomic Computing. In both concepts, the procedure ofmatching service provisioning to service demand is often centrally(geographically) realized by a dedicated resource broker instance.However, one of the prerequisites of this resource broker wouldbe that it scales with the increasing size of the system. This articleexamines the broker’s behavior with regard to a varying numberof participating nodes and shows that incremental losses have tobe accepted in central resource allocation when introducing newnodes. In comparison, a self-organizing, decentralized, bilateralbargaining approach shows more evenly behavior in responsetimes and allocation efficiency....