Reasch On System-level Fault Diagnosis Of Network System

With the scale increasing of multiprocessor systems,the reliability of system become more and more important.With this,the fault diagnosis which is used to ensure system reliability becoming more and mBecause of can find out the fault processors only by the processing ability of the ore significant.system without the help of external forces,system-level diagnosis has received extensive attention and research.In the system-level diagnosis research of the past,the achievements of pessimistic diagnosable under MM model was rare.In order to solve this problem,the relationship between the pessimistic diagnosable under MM model and the allowable faulty set(AFS)was proved,as a result,the characterization theorem of this problem was obtained.On basis of this theorem,and with the help of the method which was used to transforming MM model into PMC model,the pessimistic diagnosable algorithm MMFDA whose time complexity is O(n3)was obtained.Any more,most of system-level diagnosis research under the traditional ways was almost from the angle of take the system as a whole,which ignored the local structure characteristics.For that problem under PMC model,the concept of node diagnosis was introduced.And through the study of nodes diagnosis,we got the sufficient condition for node diagnosiability.Then,in order to use the sufficient conditions,ST structure was proposed.Based on the ST structure,the diagnosis algorithm called STFDA was put forwarded,and it's time complexity was proved to be O(n5),which was significantly reduced compared with other algorithms.Finally,in order to test the validity of the sufficient conditions,an analysis of n-dimensional hypercube network and n-dimensional star network by node diagnosis was performed.And the algorithm was applied to these two kinds of networks performed the fault diagnosis.