Research On Key Node Mining Method For Software Network Failure Propagation

Now the size of software research object becomes large,applications become more complex,multi state coupling between objects and logic structure moreloose,and the interaction between the basic entities of software systems becomes complex,causes the scale and complexity of software systems increase,it is becoming more and more difficult for developers to understand and control the structural features and faults of software systems,which seriously affected the development and maintenance of software project.The increase of software complexity directly leads to the complexity of software failure,and increases the cost of software development and maintenance process.The reliability of software is more and more difficult to guarantee because the ratio of system failure is directly caused by software failure.In testing and maintenance,it is difficult for developers to accurately deal with the problems associated with software systems.Therefore,dig out the useful data information from potential internal software source code or dynamic execution process,find the fault possibility and fault propagation ability of the key nodes,helps the staff to quickly understand the software structure of the software.It is of great value to improve the efficiency of software development and maintenance,find out the faults of software system and reduce the damage to the system effectively.Therefore,based on the research of software network modeling and characterization methods,dynamic execution trace sequential pattern mining method is applied to the software network,the key nodes mining representative software network behavior patterns and effects of software failure,failure analysis software evolution and support.The main research work of this paper is as follows:First of all,based on the structure and behavior characteristics of software,taking the software function as the research granularity,and aiming at the dependence and call relationship of software functional entities,We propose a modeling and characterization method for software networks.Abstract description of software network nodes based on function granularity can help to understand software system better.In the topological structure of software network,by analyzing the function node to obtain the relevant functional characteristics and topological information,and analyzing the call relationship between function and function,we can get a lot of valuable data.Then,using related software,we use the method of tracking software execution path,dynamically obtain the actual execution trajectory data of the original software internal function,the software execution path is transformed into the corresponding software execution sequence,through the proposed simplification algorithm SCRP,the continuous repeating pattern in software execution sequence is simplified,and the software execution sequence is pre processed,so as to construct the software network model and its mapping method.Secondly,based on software network and dynamic execution sequence,the maximum utility calculation method of software network sequence is designed.The software execution sequence is ordered,represents the internal function call relationship between software,software execution path includes many aspects,from the perspective of sequential pattern mining,including the implementation of software dynamic execution path.Software execution trace contains rich information involved in software reliability analysis.This research through the establishment of corresponding with the utility value of the index table of the UI-list data structure,and then use UI-list data structure to record the sequence of utility value and location information,and proposes a software implementation of efficient continuous mining sequence pattern mining algorithm SEHUCP-Miner.Using the index table UI-list data structure to establish the backward expansion strategy,on this basis,using the new utility upper bound model ruub,and then mining efficient software network representation model.Again,in the software in the network,as a function of component software system,the communication function between each call,which affect the connectivity of network software,plays an important role in the stability and robustness of software.The call between functions may propagate error defective data,which affects the security and stability of the entire software network.According to the characteristics of cumulative function of fault in software network,we define the node failure possibility function in software network metric NFP,according to the failure propagation characteristic function,we define the node failure propagation ability function metric NFPC.On this basis,the comprehensive function of node failure possibility and failure propagation ability,establish the function of the failure severity metric SOF,and design the function of the failure possibility of mining algorithm,find the set of nodes witch is large fault severity function in software network,and then mine the key nodes of network software.Finally,by using these software network key nodes and using fuzzy community partitioning method,we design a software evolution community outlier detection algorithm based on community matching SECOD.Using this algorithm,we calculate the community matching matrix and outlier matrix,obtain the function membership matrix of software program,and realize the fault and evolution analysis of complex software network.Accordingly,we can trace the characteristics of software failures in software evolution,and improve the ability of software developers to control software failures and control software quality.