A fault tolerant model to attain reliability and high performance for distributed computing on the Internet

The aim of the proposed fault tolerant model is to attain reliability and h igh performance for distributed computing on the Internet. The novelty of t his model lies in the integration of three unique schemes that work in unis on within a single framework. These three schemes are consecutive message t ransmission, adaptive buffer control, and message balancing. Message balanc ing essentially seeks to ensure that each message queue is served for an in terval, which depends on the current length of the queue, by the processor. In the experiments, only two parameters: current buffer length and rate of change of the actual queue length were used for proportional and derivativ e feedback control of adaptive buffer management. Test results have indicat ed clearly that the model goes a considerable way towards achieving the sta ted aim. (C) 2000 Elsevier Science B.V. All rights reserved.