End-to-end performance modeling for distributed network management systems

Abstract

A unified modeling framework for end-to-end performance evaluation in a network management system is presented. A network management application, e.g., alarm surveillance, can be mostly accomplished by several processes distributed over a network. In order to predict the completion time for such an application, components, whether software or hardware, incurring a delay must be identified. A proposed model, which is referred to as the end-to-end performance model, specifies: (1) components which incur a significant delay in a management application, (2) their characteristics, and (3) their interactions. In determining the major delay components, management application processing, protocol processing, transmission delay, and protocol conversion, are incorporated. A particular modeling procedure is explained using a SONET (synchronous optical network) management network.< >