Distributed Reconfigurable Simulation for Communication Systems

The simulation of physical-layer communication systems often requires long execution times. This is due to the nature of the Monte Carlo simulation. To obtain a valid result by producing enough errors, the number of bits or symbols being simulated must signiÞcantly exceed the inverse of the bit error rate of interest. This often results in hours or even days of execution using a personal computer or a workstation. ReconÞgurable devices can perform certain functions faster than general-purpose processors. In addition, they are more ßexible than Application SpeciÞc Integrated Circuit (ASIC) devices. This fast yet ßexible property of reconÞgurable devices can be used for the simulation of communication systems. However, although reconÞgurable devices are more ßexible than ASIC devices, they are often not compatible with each other. Programs are usually written in hardware description languages such as Very High Speed Integrated Circuit (VHSIC) Hardware Description Language (VHDL). A program written for one device often cannot be used for another device because these devices all have different architectures, and programs are architecture-speciÞc. Distributed computing, which is not a new concept, refers to interconnecting a number of computing elements, often heterogeneous, to perform a given task. By applying distributed computing, reconÞgurable devices and digital signal processors can be connected to form a iii distributed reconÞgurable simulator. In this paper, it is shown that using reconÞgurable devices can greatly increase the speed of simulation. A simple physical-layer communication system model has been created using a WildForce board, a reconÞgurable device, and the performance is compared to a traditional software simulation of the same system. Using the reconÞgurable device, the performance was increased by approximately one hundred times. This demonstrates the possibility of using reconÞgurable devices for simulation of physical-layer communication systems. Also, an middleware architecture for distributed reconÞgurable simulation is proposed and implemented. Using the middleware, reconÞgurable devices and various computing elements can be integrated. The proposed middleware has several components. The master works as the server for the system. An object is any device that has computing capability. A resource is an algorithm or function implemented for a certain object. An object and its resources are connected to the system through an agent. This middleware system is tested with three different objects and six resources, and the performance is analyzed. The results shows that it is possible to interconnect various objects to perform a distributed simulation using reconÞgurable devices. Possible future research to …