| Budget Amount *help |
¥7,400,000 (Direct Cost: ¥7,400,000)
Fiscal Year 1996: ¥1,500,000 (Direct Cost: ¥1,500,000)
Fiscal Year 1995: ¥1,400,000 (Direct Cost: ¥1,400,000)
Fiscal Year 1994: ¥4,500,000 (Direct Cost: ¥4,500,000)
|
|---|
| Research Abstract |
Recently, many research, institutes have supercomputers, However, conventional traffic simulation methods cannot take advantage of vector processing capabilities of a supercomputer. A new simulation method fit for supercomputer is proposed. In this method, simulation parameters such as call originations and link states for a fixed time, are transformed into vectors and matrices. The performance using the vectorized simulation method for circuit, packet, and ATM switching networks is compared with that of the conventional method. The proposed simulation method gave similar loss probability and delay distribution. However, the CPU time of the proposed simulation method is significantly short.
In these days, another type of supercomputer is becoming popular. Thatis parallel computer. A new traffic simulation method by using parallel computer is also studied. This method is applied to ATM network. ATM traffic has two levels, that is, call-level and cell-level. Call-level simulation, which means that we can see only whether connections are established or not, is comparatively easy. Besides this, cell-level simulation is also needed. We have to be able to see which cells have arrived and which have not. This is necessary to obtain the cell loss rate and the delay distribution. In ATM network, so many cells are transmitted and received in a short time, that the simulation needs a lot of computation. Therefore, it is difficult to make cell-level simulation of a large size ATM network by using conventional computer.
A high-speed ATM traffic simulator is developed, which works on a parallel computer with cell-level observation supported. Clustering of the nodes of the network is needed to assign tasks to parallel processors. It is optimized in order to minimize the communication time between processors. The CPU time has been considerably shortened, compared with a conventional main frame computer.
|
