An architecture for the coordination of system management services

Today, system management services are implemented by dedicated subsystems, built using proprietary system management components. These subsystems are customized to automate their operations to the extent feasible. This model has been successful in dedicated enterprise environments, but there are opportunities to broaden the scope of these services to multicustomer utility computing environments while reducing the costs of providing these services. A new model suitable for utility computing is being developed to address these opportunities. This model features several new elements: (1) a repository to represent the state of the remotely managed components and of the services that manage them, (2) a repository of policies and operational constraints, and (3) a set of meta-management services that use existing management services to analyze, construct, and safely execute a complex set of management tasks on remote systems. The meta-management services manage the system management services provided by the utility—;they guide and modify the behavior of the services, often as a result of the collective analysis of the state of one or more services. In this paper, we describe requirements and behaviors of such meta-management services and the architecture to provide them. We focus on the components of this architecture that enable and provide effective meta-management services in a utility environment.