Showing papers presented at "Central and Eastern European Software Engineering Conference in Russia in 2013"

Advanced study of SDN/OpenFlow controllers

Abstract: This paper presents an independent comprehensive analysis of the efficiency indexes of popular open source SDN/OpenFlow controllers (NOX, POX, Beacon, Floodlight, MuL, Maestro, Ryu). The analysed indexes include performance, scalability, reliability, and security. For testing purposes we developed the new framework called hcprobe. The test bed and the methodology we used are discussed in detail so that everyone could reproduce our experiments. The result of the evaluation show that modern SDN/OpenFlow controllers are not ready to be used in production and have to be improved in order to increase all above mentioned characteristics.

Static analysis for dynamic updates

Abstract: Dynamic system update problem, DSU, is about replacing a running application with another version without interrupting its work or disrupting its users. Some solutions for DSU support only certain kinds of updates, which makes it dangerous to apply them without confidence that an update succeeds. This paper formalises an approach for a static analysis of updates. We define a method of producing lists of exact changes occurred between versions of packaged Java applications and cover essential for an update system analysis queries: producing a list of all changes in an update and asking if a given dynamic update system supports an update. We introduce a multilevel update event system capable of describing generic packaging modifications as well as Java language specific constructs updates. We analyse recent releases of a few open-source Java projects to determine what changes their updates usually consist of. Described system has an extensible design that allows analysing object-oriented applications and can support different packaging schemes. The method has been implemented and tested in production environments.

DPMine/P: modeling and process mining language and ProM plug-ins

Abstract: Process-aware information systems (PAIS) enable developing models for interaction of processes, monitoring accuracy of their execution and checking if they interact with each other properly. PAIS can generate large data logs that contain information about the interaction of processes in time. Studying PAIS logs with the purpose of data mining and modeling lies within the scope of Process Mining. There is a number of tools developed for Process Mining, including the most ubiquitous ProM, whose functionality is extended by plugins. To perform an object-aware experiment one has to sequentially run multiple plugins. This process becomes extremely time-consuming in the case of large-scale experiments involving a large number of plugins. The paper proposes a concept of DPMine/P language of process modeling and analysis to be implemented in ProM. The language under development aims at joining separate stages of the experiment into a single sequence, that is an experiment model. The implementation of the basic semantics of the language is done through the concept of blocks, ports, connectors and schemes. These items are discussed in detail in the paper, and examples of their use for specific tasks are presented ibid.

GLR-based abstract parsing

Abstract: parsing is an important step of the processing of dynamically constructed statements or string-embedded languages (such as embedded or dynamic SQL). Existing LALR-based algorithms have performance issues. To increase performance we propose to use a GLR-algorithm as a base for abstract parsing and to reuse graph-structured stack and shared packed parse forest. RNGLR-algorithm modification for abstract parsing is presented.

The role of configuration management in outsourcing and distributed development

Abstract: The use of distributed development teams is becoming more common and for many good reasons. Some of the advantages are that it makes it possible to outsource parts of the development effort, gives access to a larger pool of talents and specialists, facilitates the integration of mergers and acquisitions, and allows for more flexibility in scaling up and down projects. However, distributed development also brings many new problems to be dealt with on a project. It is more complex to manage, tends to create silos between groups, and there is a risk of loss of control over remote groups or people. Traditionally Configuration Management is seen as the infrastructure that allows for the coordination of the various activities on a project and it makes sure that work products flow smoothly through different stages of the development process.In this paper, we want to investigate to what degree Configuration Management concepts and principles can provide an infrastructure for distributed development teams too and help address some of their special challenges. We first look at what challenges distributed development teams face and categorize them according to how closely related they are to the Configuration Management domain. Then we sketch Configuration Management solutions to some of the related challenges. It turns out that surprisingly many distributed development challenges can be dealt with or made less problematic by simply applying well-known concepts and principles from Configuration Management. Furthermore, a number of other challenges can be alleviated by creative thinking in the implementation of Configuration Management and/or the collaboration between the Configuration Management activity and other activities.

Evolutionary software development using procedural-parametric programming

Abstract: The article examines the methods of evolutionary software development with the use of the procedural parametric paradigm of programming. The basic program object of the paradigm are presented and their implementation in a programming language Alien is shown. The generalized records and generalizing parametric procedures are described as well as their using for the enlargement of already written a program without changing the previously written code. We consider a number of techniques that demonstrate the peculiarities of the proposed programming style. The use of generalizing procedures combined with generalized records instead of extensible types provides support for the multiple polymorphism as a special case includes a single polymorphism that is used in the objective-oriented approach. The flexibility of the proposed programming style in case of adding new data types is considered on the example of multimethods. In order to enhance the language capability it is proposed to use plug-in modules.We have carried out the analysis of the situations that arise in the software development process and compared the possibility of using evolutionary expansion in procedural, object-oriented and procedural parametric approaches. We have considered adding new specializations to the generalizations, adding new procedures, adding new data fields to existing types, adding a new procedure designed to process only one of the specialization, creating a new generalization based on the existing specializations, adding new multimethods to the program and changing multimethods.The article shows that the use of parametric generalizations combined with generalizing parametric procedures leads to the creation of more flexible evolutionary-enlargement programs. The complexity of building for generalized records is comparable with the creation of enlargement records with the use of programming language Oberon or class hierarchies of object-oriented languages. The use of plug-in modules facilitates the control of the program increment process.The development of the programming language Alien allows us to carry out the research of capabilities for procedural parametric paradigm and to define the further ways of its development. In particular, the tool support for the evolutionary enlargement of multimethods allows carrying out a flexible direct coding without additional cost in the case of degeneration of multimethods to a simple constructions. Many of these constructions are currently implemented using a pattern design. The procedural parametric paradigm in general allows us to describe many of the patterns directly, ie, without specifying complex relationships between classes and without their dynamic binding at run time. This increases the efficiency of the written code and reduces its size.

Forecasting software development project characteristics using meta-modeling

Abstract: Applying forecasting principles to software development projects is becoming one of the main trends in the modern commercial software development. There are several reasons for that, but one of the major is the rapid industry growth, which sets high expectations for the results quality and delivery timelines. Forecasting becomes possible through utilizing modern software development methodologies, simulation and statistical models.This research introduces a method to create custom software development project models (to be further used for forecasting) based on the so-called meta-modeling principles. The article defines the meta-model concept, its structure and principles, gives examples on how it can be used to simulate real software development projects.

Lightweight Linux dynamic libraries profiling technique for embedded systems

Abstract: Situations when profiling is required for embedded applications are common in software engineering practice. Usually such tasks have additional limitations - no ability to recompile or relink application, necessity to minimize profiler impact on application work, which makes performance evaluation difficult problem. The paper describes lightweight technology for solving this problem on Linux-based systems and implementation details. The key point of proposed method is a transparent integration to the dynamic linking and loading processes, which allows to profile user-defined set of functions without modification of analyzed software and with minimal impact on software and OS functionality and performance.