Proceedings PaperNet-Faim: distributed computation of aerial images
|Format||Member Price||Non-Member Price|
Simulation of aerial images is an important part of modern microchip manufacturing, but computation of the image of an entire mask is a challenging problem requiring a large amount of memory and CPU time. Fortunately, it is possible to decompose the large problem of computing the full image into many smaller, mostly independent, sub-problems. In this paper, one particular decomposition is described and implemented. The target platform is a heterogeneous group of networked workstations. The program, net-faim, was designed to be robust, to scale well with available resources, and to place modest demands on participating workstations. All of these design criteria have been realized. The overall performance of the distributed computation is linearly proportional to the sum of the performances of the individual processors, up to a rather high level of parallelism. Robustness is achieved by not relying on any one server to complete a given task; instead, if an idle server is available, the task is sent out to the idle server even if it has previously been sent to another server. The task is only retired when a server returns the completed answer. This 'paranoid' method of processing tasks has the pleasant side effect of doing automatic dynamic load balancing. The results of runs with several different configurations, both of participating workstations and of sub- domain sizes, are displayed.