利用反向傳播算法合理分配緩沖區(qū)
Managing Buffer Appropriately with the Back Propagation Learning Algorithm
-
摘要: 該文提出了一種新的緩沖區(qū)分配方法，即動態(tài)神經(jīng)共享 (Dynamic Neural Sharing，DNS)方法。這種方法利用反向傳播算法合理分配緩沖區(qū)資源，從而減少自相似業(yè)務的分組丟失率。通過兩組仿真實驗發(fā)現(xiàn)，與完全分割(Complete Partitioning, CP), 完全共享(Complete Sharing, CS), 部分共享(Partial Sharing, PS)這些傳統(tǒng)的緩沖區(qū)分配方法相比，DNS在減少分組丟失和體現(xiàn)公平性(每個源都占有一定數(shù)量的緩沖區(qū)資源)之間達到了更好的平衡。Abstract: A novel buffer management algorithm named DNS(Dynamic Neural Sharing) is suggested in this paper. This algorithm utilizes the Back Propagation learning Algorithm(BPA) to manage buffer appropriately, thus reduce the packet loss in self-similar teletraffic patterns. A conclusion is drawn through two emulations that the DNS addresses the trade off between packet loss and fairness issues better than those traditional algorithms such as CP(Complete Partitioning), CS(Complete Sharing) and SPS(State Partial Sharing).
-
Yousefi'zadeh H, Jonckheere E A, Silvester J A. Utilizing neural networks to reduce packet loss in self-similar teletraffic patterns[J].IEEE International Conference on Communications, Anchorage, AK, USA. ICC0.2003, (3):1942-1946[2]Yuang M C, Tien P L, Chen C S. A contention access protocol with dynamic bandwidth allocation for wireless ATM networks. IEEE International Conference on Communications. New Orleans, USA, ICC2000, 2000, (1): 149-153.[3]Bolla R, Davoli F, Maryni P,et al.. A neural strategy for optimal multiplexing of circuit- and packet-switched traffic[J].Global Telecommunications Conference, Orlando, Florida, USA, GLOBECOM9.1992, (3):1324-1330[4]Long T W. A self-similar neural network for distributed vibration control. Proceedings of the 32nd IEEE Conference on Decision and Control, San Antonio, Texas, USA, 1993, (4): 3243-3248.[5]Mohamed S, Rubino G. A study of real-time packet video quality using random neural networks[J].IEEE Trans. on Circuits and Systems for Video Technology.2002, 12, (12):1071-1083 -
計量
- 文章訪問數(shù): 2198
- HTML全文瀏覽量: 84
- PDF下載量: 589
- 被引次數(shù): 0
下載:
