Reliability estimation for large distributed software systems
CASCON '08 Proceedings of the 2008 conference of the center for advanced studies on collaborative research: meeting of minds
| Hi-index | 0.00 |
The exact calculation of all-terminal networkreliability is an NP-hard problem, with computationaleffort growing exponentially with the number of nodes andlinks in a network. Because of the impracticality ofcalculating all-terminal network reliability for networks ofmoderate to large size, Monte Carlo simulation methods toestimate network reliability and upper and lower bounds tobound reliability have been used as alternatives. In thisstudy, an artificial neural network (ANN) is used toestimate all-terminal network reliability for networks withboth homogeneous and heterogeneous link reliablity. Twoforms of design methods for generating training data setsfor networks with homogeneous and heterogeneous linkreliability are compared. These are experimental designand randomized design