Research On Mining Important Nodes In Software Execution Network Based On Cascading Failure

For a real-time software system, the key entities in software play an important role in software running procedure, which is closely related to software quality. As a critical part of software entities, evaluating the importance of software nodes in software running process becomes an indispensable aid in software understanding. Due to the differences of current importance evaluation methods existed in diverse conditions, the suitable one for entities importance evaluation and mining in software systems is not perfect. Thus, from the perspective of complex network to re-study software system, the nodes in software execution network are studied based on cascading failure, and the main research issues involved in the paper are as follows:Firstly, the static structure of software system can not reveal all characteristics of software system. For the sake of this, the stacks’ changes occurred in the process of software execution are tracked in this paper. Then the information of software dynamic execution process is gained, which is used to modeling the directed-weighted network of software execution.Secondly, due to the close connection among different nodes in software execution network, as one or few of them malfunction, their callers might undergo failure through the call or dependency between them and eventually crash fractional or entire software system, which would result in cascading failure and even bring about security issues. On the foundation of software execution network above, the failure propagation ability and probability of nodes are defined in this paper. And following the failure propagation strategy, a cascading failure diffusion analyzing(CFDA) algorithm is put forward to evaluate the impact of node failure on software dynamically during its runtime.Thirdly, by the aid of cascading failure method, nodes’ failure spreading range and the number of infected nodes are analyzed by malicious attacks on high in-strength nodes and strong failure propagation probability nodes in software execution networks respectively. By this, we explore what kind of nodes failure is more influential when under attack and should be given more protection.Lastly, under the circumstance of cascading failure caused by attacking nodes, the node failure influence to software execution network is evaluated to assess the importance of nodes. The corresponding experiments are conducted in this paper, and the practicability of method is also being verified by experiments.