Multi-step model for prognostication and detection of telecommunication network overload

DOI: 10.31673/2412-4338.2019.023543

Abstract

As a result of the conducted research, a scheme of multi-step prediction of the state of the queue is proposed and substantiated. The paper uses a general theory of sensitivity with indirect feedback. The results of this theory have been used to build an indirect feedback control system that saves channel and computing resources. Traditional approaches to solving the problem of congestion management are based on the control of the quantitative value of the queue length of the input and output flows of the network. The congestion identificator can only indicate that there is a congestion on the connection, but not the location or cause of the congestion. A simple increase in buffer capacity causes an increase in the number of packets destined for retransmission. This solution causes unacceptable increases in service delays.
The paper presents a new method of optimal control of congestion of a telecommunication network using the sensitivity function and dynamic neural prognostication models. The proposed algorithm is based on predicting and managing the activity of message sources. Prognostication models are designed to manage circular delays in the data transfer process. Two neural architectures are considered for predicting system output: a model with shared and separate processing of traffic with different priority. The models considered provide a further understanding of the possibilities and limitations of the proposed methods for managing neural prognostication. The proposed solutions are based on the use of advanced neural network architecture, so the important task is to develop algorithms for learning neural network for dynamic data transfer processes.

Keywords: sensitivity function, congestion, activity of message sources, neural prediction models, traffic priority, circular delay.

References
1. Kurose J. F., and Ross K. W. (2017). Computer Networking: A Top-Down Approach, 7th ed. Pearson Education, Inc., 864.
2. Keshav S. (1991). Congestion Control in Computer Networks / S. Keshav.– Ph.D. Thesis, University of California.
3. Göransson P., Black C., and Culver T. (2017). Software Defined Networks: A Comprehensive Approach, 2nd ed. Morgan Kaufmann, US, 409.
4. Shooman M. L. (2002). Reliability of Computer Systems and Networks – Fault Tolerance, Analysis and Design. JohnWiley&Sons, Inc., NewYork, 546.
5. Toroshanko Ya. I. (2016). Managing the Reliability of the Telecommunication Network Based on Analysis of the Complex Systems Sensitivity // Telekomunikatsiini ta Informatsiini Tekhnolohii, 3, 31–36.
6. Tomovich R., and Vukobratovych M. (1972). General Theory of Sensitivity. Moscow: Sovyetskoe Radio, 240.
7. Vinogradov N. A., Drovovozov V. I., Lesnaya N. N., and Zembitskaya A. S. (2006). Analysis of the Load on Data Networks in Critical Application Systems. Zviazok. 1 (61). 9–2.
8. Zhaoming Lu, Qi Pan, Luhan Wang, and Xiangming Wen (2016). Overload Control for Signaling Congestion of Machine Type Communications in 3GPP Networks. PLOS ONE, 11. DOI: 10.1371/journal.pone.0167380.
9. Toroshanko Ya. I. (2016). Analysis of Sensitivity of Mass Service Systems Based on the Model of Adaptation and Regulation of External Traffic. Visnyk Khmelnytskoho Natsionalnoho Universytetu, 6(243), 171–175.
10. Tarhov D. A. (2014). Neural Network Models and Algorithms. Moscow: Radiotehnika, 352.
11. Galushkin A. I. (2010). Neural Networks: Basic Theory. Moscow: Goryachaya liniya – Telekom, 496.
12. Stallings W. (2016). Foundations of Modern Networking: SDN, NFV, QoE, IoT, and Cloud. Pearson Education, Inc., Old Tappan, New Jersey, 544.