Most web-based simulation is aimed at modeling, particularly at building simulation languages and at creating model libraries that can be assembled and executed over the web.
At the same time, discrete event simulation has become a popular tool for designing large, man-made systems like communication networks, traffic systems, and manufacturing facilities. It offers excellent graphical animation capability and an almost unlimited degree of modeling flexibility (as a simulation model can incorporate arbitrary levels of system detail).
As technology advances, simulation models also become more complex, incorporating increased levels of system detail, higher levels of system activity, and increased system reliability. While affording greater modeling realism, execution speed has been severely strained, threatening the viability of simulation as a decision support tool.
One approach to gain execution speed is to adopt the parallel simulation technology, wherein simulation models can be executed on different (or parallel) computers where the processes are split and executed simultaneously. Traditionally, the application of parallel simulation technology has been limited, due to expensive hardware requirements and the scarcity of system software that can support large-scale distributed simulations.
Chun-Hung Chen, Associate Professor of Systems Engineering & Operations Research at George Mason University, and Enver Yücesan, Professor of Operations Management at INSEAD, explain that the Internet and web-based technologies provide a solution for the size and efficiency of parallel discrete-event simulations.
On the Internet, you can find a viable infrastructure for parallel discrete event simulation with no need for parallel hardware or networking capabilities. Just like an Internet site that can be accessed from by user with a computer and a web browser, it makes the infrastructure available to multiple users in many places (without necessarily requiring uniform user behavior). Thus, web-enabled simulation environments become feasible because of their portability, maintainability, and conformance to standards.
The objective of the authors research is to introduce a framework that combines efficiency with effectiveness: the statistical efficiency of simulation optimization techniques with the effectiveness of parallel execution algorithms. The authors introduce the Optimal Computing Budget Allocation (OCBA) algorithm, a novel simulation sampling procedure implemented in a web-based environment for low-cost parallel and distributed simulation experimentation.
OCBA works to identify the optimal system through an optimal experimental design, based on the premise that when there are several design choices, a larger portion of the computing budget should be allocated to the designs that are critical in identifying the best one. Instead of equally distributing simulation replications for different design alternatives, OCBA offers an intelligent way to allocate experiments to processors.
The key contribution of OCBA is the significant gain in simulation efficiency through significant reduction in computational effort (i.e., the total number of simulation replications needed to identify the best system or obtain a desired simulation quality). Moreover, OCBA offers a natural way of running the optimization in a distributed fashion, thereby providing further gains in simulation efficiency.
In the working paper, they offer background information on: simulation optimization (through OCBA); parallel and distributed simulation; comparison of four approaches to exploiting parallelism in discrete-event simulation to reduce computation time; and web-based simulation, explaining the advantages of using Java as the key enabler. Then they describe how web-based OCBA is implemented using a single processor, and finally they explain the prototype for distributed simulation over the Internet.