A performance evaluation tool for hybrid and dynamic distributed systems

Abstract

Distributed systems are usually modeled by a set of distributed processes spread over a number of networked computers. Such processes communicate and synchronize themselves by message passing through communication channels. Processes and communication channels can be characterized by synchronous or asynchronous timeliness behavior, according to the characteristics of underlying systems (operating system and communication sub-system). Unlike conventional distributed systems, the timeliness characteristics of dynamic and hybrid distributed systems may vary over time, according to the availability of resources and occurrence of failures. Such systems are becoming common today because of the increasing diversity and heterogeneity of computer networks and associated devices. Due to their high complexity, these systems are difficult to test or verify. In this paper, we introduce a novel simulation tool for such environments, where distinct fault models and timeliness properties can be dynamically assigned to processes and communication channels. Such a tool is meant not only for protocol evaluation but also for prototyping, allowing code reuse in real applications.