A convergence algorithm to help enhance the performance of distributed systems on large networks

The proposed convergence algorithm quickly and accurately predicts the mean message response time and communication channel. From the predicted value the correct window size of the desired percentage of time-out tolerance in message transmission and response can be computed. A correct window size is a necessity for reducing excessive message retransmissions caused by time-outs. The convergence algorithm is particularly useful for enhancing the performance of time-critical distributed applications running on large networks such as the Internet. Prolonged delays in message response in these systems may lend to fatal errors. The discussion in this paper concentrates mainly on the theory and verification of the convergence algorithm. The simulation results have confirmed that the proposed algorithm is indeed a cost/effective solution for enhancing system performance. The simplicity of the convergence algorithm makes it worthwhile for practical implementation in real systems.

[1]  Daniel S. Yeung,et al.  A proposed model for developing distributed programs by prototyping , 1998, SMC'98 Conference Proceedings. 1998 IEEE International Conference on Systems, Man, and Cybernetics (Cat. No.98CH36218).

[2]  Isi Mitrani,et al.  Probabilistic Modelling , 1998 .

[3]  Peter G. Harrison,et al.  Performance modelling of communication networks and computer architectures , 1992, International computer science series.

[4]  Bruce P. Lester The art of parallel programming , 1993 .

[5]  Jon Crowcroft,et al.  Analysis of shortest-path routing algorithms in a dynamic network environment , 1992, CCRV.

[6]  Ashok K. Agrawala,et al.  Experimental assessment of end-to-end behavior on Internet , 1993, IEEE INFOCOM '93 The Conference on Computer Communications, Proceedings.

[7]  Vern Paxson End-to-end internet packet dynamics , 1999, TNET.