Comparison of bootstrapping and finite state machine simulations of a scheduling benchmark

This paper compares simulation of a scheduling benchmark using two different approaches. The benchmark mimics the behavior of a chat application, with sender and receiver threads. One approach involves representing the benchmark as a Finite State Machine (FSM), in which the various threads could have different states, and move from one state to another until they complete. Another approach uses a benchmark performance approximation method known as bootstrapping, introduced in the AKULA scheduling simulator. The benchmarks are simulated in the LinSched Linux scheduler simulator. Our results show that bootstrapping produces results (actual runtime figures) closer to those obtained when running the actual benchmark on an operating system. The simulation also completes much faster with bootstrapping. However, the FSM results in a better model of scheduling, with higher significance.