The Research Of Network Management Software Reliability Engineering Framework And Some Key Technologies

In order to guarantee the normal, economic, reliable and secure operation of network, the reliability of network management software as the component of network management has been gaining world wide attention. As a domain application, network management software has characteristics of itself and generic software.In this paper, the framework and implement method of network management software reliability engineering are proposed based on summarizeing the research of software reliability and network management. Moreover, some concrete reliability design existing in current network management software are modeled and analyzed. The main achievements are as follows:(1) According to the research of software reliability domain, the framework proceeding reliability engineering in network management software is proposed, giving a demonstration of WDM network management software. The researches contents include the definition of failure and failure severity in network management software, the selection of software reliability model, the allocation of software reliability and the software reliability design etc. Moreover, the detailed methods are proposed, considering the characteristic of network management software.(2) A new method for analyzing module reliability of network management software in operational phase is proposed to resolve the problem existed in the domain of current network management software reliability evaluation. Firstly, some methods, e.g. periodic survival reports and module redundancy, which is used in network management software are assembled into three major run-time scenarios; Secondly, three scenario are modeled using Markov Regenerative Stochastic Petri Nets (MRSPN), and the effect which various scenario impose on module reliability is analyzed. Finally, some important reliability index, such as module availability, is concluded with applying the models. Numerical examples show that module reliability depend on both quality itself and run-time scenarios. Suitable run-time scenario can be applied in various modules according to the requirement of network management software.(3) A new method for analyzing communication reliability of network management software using Software Bus technology is proposed to resolve the problem existed in the domain of current network management software reliability evaluation. Firstly, message transmission process under Software Bus is summarized from the viewpoints of environment and queue; Secondly, this process is modeled using time based Petri nets; finally, the model is analyzed by reduction methods and matrix geometric method, and some important reliability index, such as success probability of message transmission, is concluded with applying the models. Numerical examples show that communication reliability of network management software using Software Bus technology depend on both transmission environment and service capability of queue. Model can be applied to search for bottleneck of communication reliability and improve reliability of design and configuration.(4) A new software reliability modeling method in which maintenance policy and failure correlation is combined is proposed for resolving the problem existed in failure correlation research in the domain of current software reliability modeling. According to the characteristic of un determinate maintenance policy and six mode of state conversion, Markov Renewal Process of three class points including successful run, failed run and version update is used to establish the software reliability model respectively in discrete time and in continuous time. On the base of it, layer thought is introduced into the model: modeling dependent successive software runs at each layer and maintenance policy between two layers. Finally some important reliability measurement including the distribution function of the time to failure, software reliability, mean time between failures, expected number of failures detected and expected number of defects removed are concluded by applying this model. By models comparison and analysis, the intrinsic consistency between the proposed model and other software models is validated, and the effect which delayed defects removal impose on failure correlation is analyzed accordingly.