复杂软件程序运行异常源点准确定位仿真分析

Accurate Positioning and Analysis of Abnormal Source Points of Complex Software Programs

当复杂软件各个程序同时运行时,对应的程序节点间存在较强的差异性,当前异常定位方法进行异常源点定位过程中,没有考虑到节点间的强差异性,导致异常源点定全率和定准率较低。针对上述问题,提出基于多域划分的复杂软件程序运行异常源点准确定位方法。根据软件程序节点不同异常类型的发生几率,采用K-means算法将软件程序域分割为多个子域,使发生近似异常的节点聚集在相同的子域内。提取子域内异常节点特征,通过粒子多次迭代搜索异常源点特征最优解,获得的即是异常源点定位结果。实验结果表明,所提方法 Custom定位效果有48%的提升,Elevator定位效果有22%的提升,Sgrep定位效果有39%的提升,极大地提高了定位精度。

An accurate method for locating abnormal source points in complex software program based on multi - domain partition was proposed. According to the probability of occurrence of different abnormal types in software pro- gram nodes, the K - means algorithm was used to divide the software program domain into many sub - domains, so that the nodes with approximate anomaly were clustered in the same sub - domain. Then, the characteristic of abnormal node in sub -domain was extracted. Finally, the optimal solution for abnormal source point features was itera- tively searched. Thus, the positioning result of abnormal source point is obtained. Simulation results show that the Custom positioning effect of proposed method has been increased by 48%, Elevator positioning effect has been increased by 22% , and Sgrep positioning effect has been increased by 39%. Thus, the positioning accuracy is greatly improved.