Analysis of Influence of Self-similarity to a Network Node

Apply

Analysis of Influence of Self-similarity to a Network Node

Journal Title: Jaunųjų mokslininkų darbai - Year 2010, Vol 28, Issue 3

Abstract

The article deals with the segment of distance studies’ server network characterised by self-similarity and fractality. The influence of network flow’s characteristics on throughput of a network node is assessed. Application of MMm (MM1) network model for serving distance studies’ servers is analysed. The packet of modelling of a network flow is worked out and prepared for operation; by using it, the self-similar and Poisson type network flows were analysed, influence of different flow’s characteristics on efficiency of a serving network node was assessed. The packet for network flow modelling consists of the sub-system of simulation of a network flow, the subsystem of modelling of a network node, the sub-system of control and the sub-system of input/ output. 6,776 time series were generated for the research, by indicating different queue parameters. In order to assess the influence of characteristics of the served flow in a network node on efficiency of service for a network node, every time queue was served several times with different parameters of queue length and service discipline. The results of 74,536 queues were obtained. The linear regression was applied for analysis of the results, when independent variables were as follows: intensity of network flow, the size of the network node’s buffer and the index of stability of a time queue characterising its self-similarity. The values of dependent variables (probability of queue service, probability of buffer filling and probability of packet loss) were calculated while presenting the results of queue service. While assessing results of the regressive analysis, the intensity of the flow was regarded. It was found out that when intensity of the network flow increases, the self-similar flow is being stably served, and probability of packet loss in Poisson type network flow increases. Probability of serving a self-similar flow increases when a service queue of a network node increases, and the buffer length of Poisson type time series makes no influence on quality of service. The discipline of service of time series LIFO ensures better quality of service when the incoming flow is of Poisson type and the serviced one – self-similar. When the stability index of a self-similar queue increases, the probability of service increases. A stronger dependence is observed when the incoming and served flows are self-similar.

Authors and Affiliations

Liudvikas Kaklauskas, Leonidas Sakalauskas

Keywords

self-similarity GI\G\1 Poison flow α-stable process FIFO LIFO Regression Analysis

Calculation of Stress-strain State Parameters at Rectangular Cross-sections of Flexural Reinforced Concrete Members According to En-2 Curvilinear Concrete Stress Diagram

EP ID EP118986
DOI -
Views 69
Downloads 0