Simulation-based analysis of performance dynamics of distributed applications in heterogeneous network environments

Next-generation computation-intensive collaborative applications in various science and engineering fields feature largescale distributed computing workflows. There exist many distributed algorithms for task scheduling,modulemapping, and service provisioning to support the execution of such workflows in heterogeneous computing environments and optimize their network performance for fast user interaction or smooth data flow. However, deploying large-scale distributed applications in real network environments is extremely challenging because of the time-varying dynamics in the reliability, availability, capacity, and capability of massively distributed system resources,which are typically shared among a broad community of users over Internet or dedicated connections. We propose a simulation system to study the execution dynamics of distributed applications and evaluate the network performance of scheduling or mapping solutions before reallife deployment and experimentation. The proposed system visually illustrates the dynamic execution process of complex workflows in network environments by simulating data communication over transport links and data processing on computer nodes in a highly distributed and parallel manner. We implement the simulation system and conduct extensive testings on various distributed mapping algorithms. The theoretical results obtained by rigorous performance analysis based on well-definedmathematicalmodels are evidently confirmed by the simulation-based performance measurements.