Object instrumentation for distributed applications management

Abstract

The development of complex distributed applications running on heterogeneous computer networks raises the strong demand for effective observation and control, i.e., for the management of the entire system behavior. Distributed systems create a new class of requirements especially in the areas of installation, consistent configuration, fault detection, and management exceeding what is necessary for stand-alone applications. In this context more flexibility and generality is required for both the management system and the instrumentation of the application components than that provided by existing network management systems. This paper proposes a method to support the development of manageable distributed applications based on a formal management interface definition and an instrumentation library. This method is also applicable to already existing components of distributed applications provided their source code is available.