ABSTRACTS OF ARTICLES OF THE JOURNAL "INFORMATION TECHNOLOGIES".
No. 5. Vol. 22. 2016

To the contents

À. Å. Saak, Associate Professor, Head of department, e-mail: saak@tgn.sfedu.ru, Southern Federal University

Scheduling of Sets of Hyperbolic-Type and Circular-Type Tasks in Grid-Systems

The annotation: Users' growing demand for computational capability, along with the progress in technologies, stimulate the transition from metacomputing to Grid-computing at the end of past century, which is still stepping up today jointly with such distributed computing paradigms as cloud computing and jungle computing. This paper considers computer tasks with a priori defined time of execution, in which the number of required processors is specified by a user at the moment of their delivery to the system. In these conditions when assigning sequentially numbered processors for task handling, the task of schedule generation by a scheduling algorithm is equivalent to the task of resource rectangles distribution. User's task, which needs service from Grid system's scheduler, is presented as a resource rectangle with horizontal and vertical dimensions, correspondingly, are set equal to the number of processors and time units required to complete a task. Grid-systems with centralized structure of their scheduling system which consist of sites which contain parallel systems which function in the mode of multisite scheduling are simulated by the resource quadrant. The author defined the resource rectangles environment, as a theoretical base of algorithmic support of polynomial complexity scheduling for the purpose of distribution management of computer and time resources and users' tasks in Grid-systems. In the environment of resource rectangles versions of level algorithms by height and by length are suggested. They allow scheduling by the sets of resource rectangles which don't have the feature of monotony of horizontal and vertical dimensions. This paper poses a question of suggested polynomial algorithms adaptiveness for the sets which consists of tasks of hyperbolic and quadratic, and circular and quadratic types. The comparative analysis allows recommending some of the algorithms to use in Grid systems with centralized structure.
Keywords: Grid system, centralized structure of scheduling system, multisite mode of task service, non-preemptive task service, circular-type task, hyperbolic-type task, Non-Euclidean heuristic measure, polynomial complexity of an algorithm, level algorithm by height, level algorithm by length

P. 323—332

A. S. Mokhov, Assistant, e-mail: asmokhov@mail.ru, V. O. Tolcheev, Professor, e-mail: tolcheevvo@mail.ru, NRU "ÌPÅI"

Approaches to Considering Patterns of Scientific Documents in Processing and Analysis of Text Information

The procedures of processing and analysis of text information, based on document structure are considered in the article. The main models of text document representation are reviewed. The effectiveness of partly structured models is shown for such tasks as information retrieval, automatic annotation and fuzzy duplicates identification. Along with the known approaches the new authors' methods are presented. These methods successfully consider distinctive features of bilingual bibliographic documents and decrease classification error.
Keywords: Text mining, text document representation, informational retrieval, automatic annotation, fuzzy duplicates identification, classification of bilingual bibliographic documents

P. 332—339

A. R. Mukhutdinov¹, Professor, e-mail: muhutdinov@rambler.ru, Z. R. Vahidova², Associate Professor, e-mail: MRZulphiya@rambler.ru, M. G. Efimov¹, Graduate Student, e-mail: jero07@bk.ru
¹KNRTU,
²KNRTU-KAI

Computer Modelling of Brisant Action of Explosion

In this article the developed and fulfilled technique of creation of computer model of brisant action of explosion for sinking of the lead cylinder (Ness's test) is presented to ANSYS AUTODYN. Comparative studying of results of modeling and experiment is carried out: shattering effects of secondary explosives of the increased power. It is shown that the computer model allows to predict brisant action of explosion with a mistake to 7 %; dependences of shattering effect of trotyl on density of its charge. It is shown that the computer model allows to predict brisant action of explosion (Hess's test) with a mistake to 3 %. It is established that character of the curve received on the basis of computer model differs from the curve constructed on experimental data which fixes sharp deviations (for 17%) in the range of density of 1,3...1,5 g/cm³.
Keywords: computer model, shattering effect, operational parameters, explosives

P. 340—343

S. A. Inyutin, Professor, e-mail: inyutin_int@mail.ru, Moscow Aviation Institute (Nation Research University) (MAI)

Method Calculation Quantitative Characteristic Computer Modular Value

Developed iterative method for computing quantitative characteristics from wearing for computer modular data formats in a parallel re-configurable computing systems. Quantitative characterization earmarked for implementation non-modular operations in a parallel computer arithmetic. The iterative method allows to obtain a characteristic with the use of additional equipment with a minimum width while the quadratic computational complexity of the calculation algorithm.
Keywords: reconfigurable multiprocessor systems, modular computing process, complexity of computation, figure characteristic for modular computer formats, numerical characteristics, cellular algorithm of the routing

P. 343—347

V. A. Bogatyrev, Professor, A. V. Bogatyrev, Graduate Student Saint-Petersburg National Research University of Information Technologies, Mechanics and Optics (University ITMO) Kronverkskiy Prospekt 49, 197101 St Petersburg, Russia, e-mail: vladimir.bogatyrev@gmail.com

The Model of Redundant Service Requests Real-Time in a Computer Cluster

Considered computing cluster architecture, operating in real time. The aim of the research is increasing the probability of timely and error-free service. This goal is achieved due to the fact that the system provides the direction of backup service requests to multiple cluster nodes with the possibility of destruction after exceeding a certain threshold expectations. Consideration of options for dispatching with destruction and without destruction of backup copies of requests waiting in the local queues of nodes in excess of the allowable time. The dependence of the probability of timely and accurate service requests from the multiplicity of backup copies of requests to the service at different nodes. Shown a significant increase in the probability of timely services as a result of reservation requests. It is shown that the scheduling queue with the destruction of expired queued backup requests can significantly improve the efficiency of service requests. The existence of effective reservation requests depending on scheduling the rate and permissible delay in queue.
Keywords: model, reliability, a redundant service requests, real-time cluster scheduling

P. 348—355

S. Yu. Mikova, student, e-mail: sofya_mikova@mail.ru, V. S. Oladko, Ph. D., Associate Professor, e-mail: oladko.vs@yandex.ru, Volgograd State University, Volgograd

Comparison of Algorithms to Identify Network Anomalies Using Measures Van Rizbergena

The problem of selecting algorithms for detecting network anomalies viewed in the article. The aim of this study is to develop an approach to the selection of the best algorithm to detect anomalies in network traffic; this approach will reduce the risk of possible attacks. The authors as a criterion for the selection of the algorithm used the criterion of the quality of the classification of network traffic. Functional model of quality research algorithms to detect network anomalies offered. As a basic measure of quality used Van Rizbergena. Program complex that automates the proposed model developed. The objective of the study is to generate network traffic with different characteristics both normal and abnormal, to assess the quality of existing algorithms for detecting network anomalies. Experimental evaluation of the quality of the two algorithms — Brodsky—Darhovsky algorithm and network anomalies detection algorithm based on discrete wavelet transform performed. These estimate algorithms showed that the best quality classification algorithm has Brodsky—Darhovsky algorithm.
Keywords: accuracy, completeness, false positive, false negative, Brodsky—Darhovsky algorithm, discrete wavelet transform, attack

P. 356—361

S. M. Avdoshin, Ph. D., professor, Head of Software Engineering School, Faculty of Computer Science, HSE, e-mail: savdoshin@hse.ru, A. V. Lazarenko, Undergraduate Student, Software Engineering School, Faculty of Computer Science, HSE, e-mail: avlazarenko@edu.hse.ru, National Research University Higher School of Economics (HSE)

Tor Users Deanonymization Methods

Tor currently is the biggest anonymous network in the world, which is not only the biggest but the safest one. That is why it is widely used by different criminals such as: killers, drug dillers, carders, etc.. The problem of users deanonimizaiton is considered as a very hard and sophisticated task, because of Tors' design and anonymity-features. However, it is possible to solve this problem. This paper is an overview of Tor users deanonimization methods. It describes passive and active methods, their key features. Passive methods are methods, where an adversary can only analyze traffic in the network, but modifications of it are restricted. Active methods actively use traffic modification for its own purposes. The majority of attacks uses Tor vulnerability, related to end-to-end traffic eavesdropping. In the modern Tor network such an attacks could be successful with a very small probability, because an attacker must control both entry and exit tor nodes of a victim. The probability of such combination is very small. On the other hand, there is a class of attacks, known as website fingerprinting attacks. This attacks are very powerful, if an attacker is smart enough in the field of data science. Moreover, WF attacks requires a particularly small amount of resources for successful usage in the real network. But if an attacker is a global adversary of the network, simple statistical tests will help him to successfully deanonimize all users in the network.
Keywords: anonymous network; deanonymization; PoS; timing attacks; Tor; traffic analysis

P. 362—372

A. S. Korovin, Student, e-mail: akorovin00@gmail.com, I. P. Skirnevskij, Assistant, e-mail: igorskir@yandex.ru, National Research Tomsk Polytechnic University, Tomsk

Web-Application for Real-Time Big Data Visualization of Complex Physical Experiments

This article examines a tool that provides the flexible user interface for representing status of various physical experiments using different visualization approaches. These experiments produce large volumes of heterogeneous data in real-time, therefore the data processing procedure is complicated. We implemented the tool as a web-based application due to capabilities of modern browsers for interactive visualization, regardless of the operating system and device type (mobile, desktop). Using the architectural style REST simplified component implementation, improved the effectiveness of performance tuning. Frequent changes m experiment conditions require the method for adapting the visualization screens, so we created the graphical user interface builder to arrange graphical control elements using a drag-and-drop technology and WYSIWYG editor. The designed system is a step in the development of applied information technology systems and allows specialists from different fields of science to analyze heterogeneous data of physical experiments effectively.
Keywords: visualization system, big data, data monitoring, client-server architecture, flexible interface

P. 373—378

U. O. Ageeva, Pupil, V. G. Ageeya, Teacher, School ¹ 1 Saratov Region, A. B. Barskiy, Professor, e-mail: arkbarsk@mail.ru, Moscow State University of Railway Engineering (MIIT)

Bionic Intelligence Limbs Prosthetic and Logical Neural Networks

Analyzes the history of prostheties and advances inprostbetic of limbs based on the bio-mehatroniks. It emphasizes the importance of solving the problem of the direct connection of the prosthesis to the brain. For this proposed use signals formed on the truncated nerve endings. Recognition and conversion of these signals into movement of the prosthesis is offered to make with the help of an artificial single-layer logical network. We consider the approximate form of a network and the activation function of the neuron. A neural network is described by the following matrix, which provides a low complexity of processing — calculation of the value of the excitation of neurons, as well as the ability to use built-in signal processors as neurocomputers. These microprocessors serve as signal amplification, and control settings. The value of the excitation of the neuron is a parameter that determines the amount of force with which the movement is performed, at which the neuron indicates. The composition of movements triggered neurons determines the resulting effect. The possible layout of the prosthesis considering reading signals from the stump, its power supply and configuration of the neural network.
Keywords: prostheties, bio-mechatronics, logical neural network, activation function, neurocomputer, intelligent prosthesis

P. 379—386

A. E. Prasolova, Chief of the Department of Economics, Computer Science and Mathematics, e-mail: prasolova@bsu.edu.ru, Stary Oskol Branch of Belgorod National Research University, Stary Oskol, Belgorod region, Russia

Collective Newral Network Algorithm for the Diagnosis of Myocardial Infarction

The article describes a neural network diagnostic system of myocardial infarction. The application of Hamming neural network, RBF network, Kohonen self-organizing maps, and collective neural network algorithm in the diagnosis of myocardial infarction using only ECG data.
Objective — to evaluate the sensitivity and specificity of the collective neural network algorithm in the problem of diagnosing myocardial infarction in a small amount of the specified parameters. Compare the experimental data with the results obtained for each neural network algorithm, which is a structural part of the alogoritma, as well as with the known results of such studies. The results of computational experiments are represented. This work describes the use of neural networks of various configurations in the diagnosis of myocardial infarction by using different localization only ECG data. We considered the following 6 parameters: the size of the Q-wave, the presence of the R-wave, the value of QRS interval, the position and the slope of the ST interval relative contour, size and position relative to the T wave contour. In a retrospective study of patients examination cardiology department with clinically confirmed myocardial infarction with different depths and localization (40 pers.) And patients who according to a survey of myocardial infarction was excluded (40 pers.) were used. Thus, the volume of the control sample consisted of 80 people.
Keywords: neural network, diagnosis of myocardial infarction, collective neural network algorithm, Hamming network, Kohonen network, RBF network

P. 386—390

V. M. Grinyak, Professor, Far Eastern Federal University, e-mail: victor.grinyak@gmail.com, Yu. S. Ivanenko, Student, Vladivostok State University of Economics and Service, A. S. Devyatisilny, Professor, Institute of Automation and Control Processes, FEBRAS

Data Rate in Distributed Information System and its Effect on Digital Radar Measurement Error

Current paper is about evaluation of measurement errors at distributed digital radar system with limited network bandwidth. Radar system considered under this paper consists of 2 main blocks: ADC convertor that digitalize raw radar signal and block of secondary data processing. Since high sampling frequency generates big digital datasets that should be transmitted from ADC to secondary processing block(s) bandwidth has crucial impact on system accuracy. Authors emphasize their efforts on analysis the systems where secondary processing block is based on common widely available hardware. Typical personal computer as common hardware is considered as suitable for research purposes. This paper is emphasis on research of system limitation based on two-coordinate radar. Problem is states as following. ADC generates amplitude matrix by signals from radar. Rows of the amplitude matrix represent angular sampling and columns represent distance sampling. The matrix is transmitted to secondary processing block that is responsible far tasks like: recognizing objects, objects tracking and rendering. The paper refer to comparison of data transmission protocols between 2 physical hardware (PC based) connected into network. Following protocols are considered as potentially suitable for inter-process communications: DDE, Mail-slots, Named-pipes, Sockets, memory-mapped files. In this paper preference is given to named-pipes. Name-pipes have been used for all computational experiments of data transmitting speed evaluation. Computational experiments for the article have been performed for system of 2 personal computers connected to local network. In order to simulate different external conditions following types of tests have been performed on the system: both hosts running data receiving-transmission without extra load, both hosts running data receiving-transmission with extra CPU utilization, both hosts running data receiving-transmission with extra memory utilization, both hosts running data receiving-transmission with extra network utilization.
The paper presents the results of data transmission experiments on charts. For 1 Gbit/s Ethernet, average transmission speed is about 50 Mbyte/s with 5—10 % fluctuations rate. One of the results of the experiments that data transmission speed does not directly depend on CPU, memory, network utilization. Based on the experimental values of data transmission speed as main result of the paper given charts of dependency of maximum size of amplitude matrix and measurement error for typical radar system.
Keywords: distributed information system, radar, digital signal processing, data transfer, measurement error

P. 391—396

R. Yu. Skorobogatov, Postgraduate, Assistant of the Department CAD, e-mail: ro-m-a-n@yandex.ru, Federal State-Financed Educational Institution of Higher Professional Education "Siberian State University of Telecommunications and Information Sciences", Novosibirsk

Introduction Virtual Characters in the Space of TV Studios

The article describes a technique for introducing a virtual character in a virtual television studio (VTS) with overlapping movements of real actors. Methods of creating video with a three-dimensional avatar. It is shown that the second generation Kinect has significantly better specifications compared to the first generation. The parameters of the equipment that meet modern requirements of the industry body.
Keywords: Kinect, 3d, virtual TV studio, rendering, virtual reality, augmented reality.

P. 396—400

To the contents