The Study Of Reliability

Novel interconnection network is the development direction of the network topology of multiprocessor computer systems. Novel interconnection network has the better properties in terms of symmetry, regularity, recursive structure and smaller diameter which indicates novel interconnection network maintains several attractive advantages including lower hardware cost, high fault tolerance, and simple routing. The reliability of novel interconnection network is the key problem which impedes the application and popularization of multiprocessor computer systems based on novel interconnection network. It is well-known that connectivity and diagnosis are two core aspects in the study on reliability of novel interconnection network. However, the reliability of novel interconnection network has not been systematically studied, so far. Some aspects have not been involved, especially in the aspects of connectivity and diagnosis, which limit the application and promotion of novel interconnection network. In this paper, we adopt some methods, such as set theory, graph theory, computer algorithm and data statistic analysis, to study the reliability of novel interconnection network. The research of this paper will help to clarify the reliable performance of novel interconnection network, and provide important theories and parameters of high reliability network in our country.In this paper, firstly, we take conditional t-diagnosable as our study object. This paper try to take advantage of F, T, M, HF four related vertex sets to study conditional t-diagnosable under the ex-test PMC model. Then we introduce several rules based on those four related vertex sets and set up a conditional t-diagnosable algorithm based on ex-test PMC model. Based on this result, we study the method of conditional t-diagnosable decision. After derive the necessary and sufficient condition of conditional t-diagnosable under the ex-test PMC model, we introduce a conditional fault sets determination algorithm which can compute all the conditional fault sets rapidly and a new conditional t-diagnosable decision algorithm. And then we focus on distinguishable measures of conditional fault sets with a new perspective on establishing functions. In terms of the constructing distinguishable function, we study how to determine a system is conditional t-diagnosable or not under the ex-test PMC model. Furthermore, a algorithm is given which shall derive conditional diagnosability. These research results have important theoretical value and practical significance to improve the diagnosis speed, reduce the calculation difficulty and improve the decisive method of conditional t-diagnosable.Secondly, we take topological properties, connectivity and diagnosis of exchanged crossed cube (ECQ) the new topology research result of novel interconnection network as our study object. After topological analyzing, we introduce some important topological properties of ECQ(s,t), then show the conditional and restricted vertex connectivity of ECQ(s,t) are 2s for t≥s>2 based on those topological properties. And then, we determine the conditional diagnosability of ECQ(s,t) under the PMC model is 4(s-1)+1 for t≥s>2. The results of the connectivity and diagnosability of ECQ will be helpful to the evaluation of reliability in the future.Thirdly, we extend the theory of g-good-neighbor conditional t-diagnosable and present a new measure of diagnosis which is called strongly g-good-neighbor conditional t-diagnosable. Besides, we research the properties of strongly g-good-neighbor conditional t-diagnosable and show the strongly g-good-neighbor conditional diagnosability of hypercube under the PMC model is lower than 22g(n-2g)+22g-1,n≥5, 1≤g<|n/3|. The strongly g-good-neighbor conditional t-diagnosable expands the boundaries of diagnosis and also enriches the diagnostic theory. It is valuable to be researched in the future.Finally, we summarize our research work, and discuss the further research work comprising the study of diagnosis under the comparison model, the study of pessimistic diagnosis, and the study of intermittent-fault and local diagnosability.In this paper, we aim to study the reliability of novel interconnection network. Through topological analysis, statistical calculation, data analysis and theoretical proof we derive several important theorems about conditional t-diagnosable, conditional and restricted vertex connectivity of exchanged crossed cube, and conditional diagnosability of exchanged crossed cube. We prove those theorems by theoretical and algorithm analysis. The research results of this paper can help to promote the application of novel interconnection network and have a strong forward-looking attractive.