Performance analysis of a multiserver discrete-time queueing system with correlated batch arrivals for ATM-based B-ISDN.
Item
-
Title
-
Performance analysis of a multiserver discrete-time queueing system with correlated batch arrivals for ATM-based B-ISDN.
-
Identifier
-
AAI9732929
-
identifier
-
9732929
-
Creator
-
Hussain, Arshad.
-
Contributor
-
Adviser: Mohamed A. Ali
-
Date
-
1997
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Engineering, Electronics and Electrical
-
Abstract
-
Most queueing analyses performed in literature are based on characterizations of queueing phenomena in continuous-time terms. Recently in telecommunication industries, Broadband Integrated Services Digital Network (B-ISDN) has received considerable attention for its capability of providing a common interface for future communication needs including video, data and voice. Since information in B-ISDN is transported by means of discrete units of 53-octets Asynchronous Transfer Mode (ATM) cells, interests in discrete-time systems have increased.;This research dissertation presents general discrete-time analytical models for the characterization of the queueing performance of an ATM network in terms of the overall performance characteristics such as buffer occupancy (in cells), jitter distribution and its impact from spatial correlation of the mono-media streams. Multistream arrivals within the same time slot are assumed to be generally correlated batch processes, but batches are independent from slot to slot. Traffic generated from mono-media sources in a multi-media session, such as voice and video, are of a stream type, and tend to be correlated. Lack of independence among mono-media sources in a multi-media environment, and the related temporal characteristics embedded in the behavior of mono-media sources, introduce cross-correlation among various mono-media streams. These correlations if not properly taken into account, as this dissertation shows, can impact network performance, such as jitter and synchronization.;The main characteristics of the present treatment lies in introducing two critical new features crucial for the accurate assessment of the queueing performance of an ATM network with multi-media traffic handling capabilities. First, unlike most of the conventional analysis which assume a 1-D known arrival process, e.g., MMPP, BMAP, DBMAP, etc., this dissertation presents 2-D general independent{dollar}\sp1{dollar} (GI) arrival process models. Second, the model assumes multiple correlated batches of information units, arriving in the same time slot, enter multiple queues with multiple servers. The media servers are of different types, such as audio server, and video server. Specifically, spatial correlations present in multi-media traffic are taken into account and numerical examples showing the overall performance of the proposed queueing model are presented. These results are then compared with that of the conventional single queue models, i.e., GI-D-c. This dissertation has established the fact that GI-D-c model and the like under or overestimate the average system contents depending on the degree of correlation among the mono-media streams and the server configurations.;Several experimental measurements in multi-media videoteleconference sessions are performed to explore and characterize the spatial correlations among mono-media sources. All experiments were performed in a Local Area Network (LAN) employing CSMA/CD. In all the experiments described in this dissertation, such correlations are demonstrated. It is envisioned that such correlations would also exist in Wide Area Networks (WAN) such as B-ISDN/ATM. Conclusions are reached regarding the influence of spatial correlation on ATM network performance by employing the experimental data into the proposed two-queue analytical model. ftn{dollar}\sp1{dollar}The numbers of messages entering the system during the consecutive time-slots are assumed to be independent and identically distributed (i.i.d) non-negative discrete random variables with an arbitrary (general) probability distribution.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
