Analysis Of Function Call Network In Software Executing Based On Complex Networking

The crossover study of software engineering and the theory of complex networkshas just started. Some achievements have been made, but most of the studies are stillneeded the more discovery, analysis, interpretation and measurement.And theseresearches have not reached the purpose of effective intervention and regulation in theprocess of software development and maintenance.The main difficulties and how toreasonable map software structure to software network, how to find some statisticalproperties of software network from a holistic perspective, how simply and effectivelyfind important nodes in the network and research on local connections mode betweennodes in software network etc.In view of this, this article will combine the theories of complex networks andrelated research method in the field of software engineering with analysis object of thereal open source software. Researches are as follows:Firstly, we track the process of software executing, and define function as anode.In the process we call relationship of function as an edge, and quantify functiondependence by callingl number, that is set to a value. This article proposes a mode ofconstructing software executing directed-weighted network. In order to understand thesoftware network better, we consider the function calls, define weighted out degree,weighted in degree, shortest path and integration.Secondly, in order to find important nodes simply and effectively in the network,a novel method is proposed to mine the functions which may widely spread failuresbased on cascading failure. We define fault infection rate (FIR) to measure theimportance of nodes in the network and propose MIN algorithm to measure and rankfunctions. MIN algorithm traverses each node in whole network and calculates Ninfectthat is the number of nodes which can arrive at the failure nodes through one or moredirected-edges. Then FIR of node is ranked to mine important nodes.Again, based on the idea of passing score, this paper put forward a method ofmeasure the importance of directed weighted network nodes. We define relative node measurement score (RNMS) for measuring the importance of functions. Node_Scorealgorithm is put forward for accurately measuring the importance of node. In order toguide developer to provide more targeted protection and defense while it develop nextversion, we predict the importance of nodes in various versions, process of upgradingof software is tracked and distribution of RNMS in each version is analyzed.Finally, using the motif mining method of biological networks and stock network,we propose a method of motif analysis method based on software executing networkto study the functions calling mode.