LOAD SHARING THAT SUPPORTS FAULT-TOLERANCE IN A DISTRIBUTED COMPUTINGSYSTEM

A scheme for combining load sharing and fault tolerance in a distribut ed computing system is proposed. Two performance models are constructe d to evaluate its performance: an approximate Markov chain model and a more detailed simulation model. The agreement between these two model s shows that the approximations in the Markov model do not affect its accuracy substantially. The overall performance results show that a di stributed system using this integrated scheme has a better performance than a system without fault tolerance or load sharing, showing that t he costs of implementing the fault tolerance are outweighed by the per formance advantages of the load sharing.