Reliability-oriented system design and analysis on computer networks

Distributed task allocation plays an important role in designing a reliable distributed system. Some previous work concerning the reliability evaluation of distributed systems showed that the distribution and redundancy of programs and data files affect the reliability of a distributed system significantly, and hence the reliability-oriented distributed task assignment has become a widely discussed issue. Some previously proposed algorithms showed that suboptimal solution can be obtained with relative error being less than 0.05 and average absolute error being 0.02. This paper proposes a reliability-oriented distributed task allocation algorithm to cope with some problems of applying previous developed algorithms. In the new approach, not only the reliability of each node and link in the network is concerned, but also the physical system constraints are dynamically taken into consideration. From the experimental results, it is concluded that desirable performance is achieved with average absolute reliability improvement A_IMP_ave=0.091 and average relative reliability improvement R_IMP _ave=13.41% compared with previously developed work