A network for parallel searching
- 128 Downloads
This paper shows a pragmatic approach to highly parallel data base machines focusing on search operations. A system of n processors connected as a superimposed tree will process as many requests in about twice log2n transfer time between processors. Since it is not difficult for a powerful microprocessor to transfer a request in a couple of milliseconds, a system of 1024 processors will have a processing rate of up to 20K search operations per second.
KeywordsMemory Unit Throughput Rate Output Stream Search Operation Calendar Data
Unable to display preview. Download preview PDF.
- 1.J. B. Dennis and D. P. Misunas, “A Preliminary Architecture for a Basic Data-Flow Processor,” Proc. 2nd. Ann. IEEE Symp. Computer Architecture, 1975, pp. 126–132.Google Scholar
- 2.H. T. Kung, “Why Systolic Architectures,” Computer, Vol. 15, No. 1 (Jan. 1982), pp. 37–46.Google Scholar
- 3.M. Maekawa, “Optimal Processor Interconnection Topologies,” Proc. Eighth International Symposium on Computer Architecture, Minneapolis, Minnesota, May 12–14, 1981.Google Scholar
- 4.H. S. Stone, “Parallel Processing with the Perfect Shuffle,” IEEE Trans. Computers, Vol. C-20, 1971, pp. 153–161.Google Scholar