A reconfigurable Markov chain simulator for analysis of parallel systems

Abstract

Markov chain is a convenient tool to analyze parallel systems for architects who are not experts of theoretical analysis. However, it is sometimes difficult to use especially when the model becomes complicated or extremely small probabilities are used in the model. In this paper, we propose a reconfigurable Markov chain simulation system and evaluate on a reconfigurable testbed. In this system, a user describes the target system in a dedicated description language called "Taico". The description is automatically translated into the Verilog-HDL description, of the Markov chain simulator. Then, the simulator is implemented on a reconfigurable testbed called FLEMING, and executed directly. From the evaluation with analysis of example parallel systems, it appears that the simulation speed of proposed system is more than hundreds times faster than software simulation on a high speed workstation.