摘要
本文提出了具有纵横输入（ＣｒｏｓｂａｒＩｎｐｕｔ）互连方式和输入缓冲（ＩｎｐｕｔＢｕｆｅｒｅｄ）结构的递归Ｋｎｏｃｋｏｕｔ交换网络（ＣＩＢＲＫＳ）．通过采用纵横输入互连方式可减少内部小交换单元的数目，并可使信元传送顺序不会受群输出端口数目的影响．而通过在每个输入端放置缓冲器可在保持丢失率性能不变的情况下，可使整个交换网络的级数减少，从而也就减少了信元在群网络中的传输时延．另外，在该结构中，通过把信元滤址的功能从每个小交换单元中提取出来放在每个输入端口，又进一步减少了小交换单元的功能．通过比较，我们认为，作为大规模ＡＴＭ交换网络结构，ＣＩＢＲＫＳ结构比传统的ＲＫＳ结构具有较好的性能／复杂度特性．
In this paper,an improved recursive knockout switching (RKS) fabric with the crossbar input interconnection and input buffered architecture (CIB RKS) is proposed.In this architecture,the number of switching elements (SWEs) is decreased and the contradiction of the cell transmission sequence and the cell loss performance measure are eliminated due to the crossbar interconnection structure.The number of the switch stage and the delay of cells through the switch fabric are decreased resulting from input buffers.In addition,the function of each SWE is simplified by separating the cell address filter with the cell concentration.We conclude that CIB RKS architecture is better than the traditional RKS architecture from the view point of the performance/complexity characteristics.
出处
《电子学报》
EI
CAS
CSCD
北大核心
1999年第4期11-14,18,共5页
Acta Electronica Sinica
关键词
ATM交换网络
B-ISDN
输入互连方式
缓冲结构
Large scale ATM switching fabric,Recursive Knockout switch,Crossbar input interconnection,Input buffer,Performance
