Introduction

Simulation models are commonly used to evaluate the performance of parallel architectures and predict the impact of algorithmic and architectural innovations on the performance of parallel applications. To reduce simulation times, most simulators typically use direct execution to simulate the local portions of their code and use models only for the communication and input/output events. Even with direct execution, sequential simulation of large parallel programs can be very time consuming [BDCW91,DGH91,CDJ91]. This has lead to a variety of attempts to use parallel execution to reduce simulation times for models that simulate parallel programs [LW96,RHL93,DHN94]. Most of the existing parallel program simulators are used to evaluate the performance of the memory hierarchy, interconnection network, or processor architecture. To the best of our knowledge, none of the existing parallel simulators have been used to evaluate parallel I/O systems. Specific contributions of this paper include:

• a simulator to predict the performance of MPI-IO programs as a function of architectural characteristics (e.g., number of I/O nodes, message passing latency), file system characteristics (e.g., data caching and partitioning), and device characteristics (e.g., seek time, data transfer time).

• performance of the parallel execution of the simulator for three benchmark applications including the BTIO benchmark from the NAS benchmark suite [BHS95]

• performance of the benchmark applications as a function of alternative implementations for collective I/O operations including two-phase I/O [RBC93]

• performance of the benchmark applications as a function of alternative caching strategies including cooperative caching [DWAP94]

The rest of the paper is organized as follows: the next section presents an overview of the subset of MPI-IO and the parallel file system that is simulated. Section 3 describes the architecture of the simulator and provides a brief desription of each of the primary components in the simulator. Section 4 describes the benchmarks and the experimental results.