Commentary—Practical Parallel Discrete Event Simulation

The feature article addresses issues that are now of crucial importance to this field. Model development for efficient parallel execution is too difficult. Our initial goal at Jade was to create a parallel simulation environment that could be used to solve a wide variety of practical problems. Fujimoto's paper outlines many of the reasons why we have not yet achieved this goal. Our experience at Jade supports Fujimoto's statement that it typically takes “highly skilled experts” to develop models for real applications that can achieve significant speedup due to parallelism. Most of the reasons he presents for this unfortunate situation also agree with our experience. However, we believe there are some important differences in how these difficulties can be addressed. An approach to the design of efficient parallel models is sketched below. This discussion identifies four key indicators of performance and suggests adding one more silver bullet to Fujimoto's list: tools that support the incremental developmen...