Open Collections

Integrated congestion management at the user-network interface of an ATM/B-ISDN network

University of British Columbia

This thesis presents an integrated congestion management platform of user traffic at the UNI of the ATM-based network considering the presence of signalling traffic. Integrated congestion management dictates that congestion control schemes are applied during the call access phase (call admission control scheme) and the information transfer phase (buffer control scheme) of user traffic source. The congestion control schemes are devised to meet the congestion performance requirements and to optimize the performance if possible. UNI call admission and buffer controls developed for the conventional packet-switched network are not applicable to the ATM-based network because of the different input traffic characteristics. In most of the past investigations on the performance of conventional packet-switched networks, the individual input traffic is mostly computer-to-computer data; such individual and aggregate traffic are well-known to follow the Poisson process. On the other hand, ATM-based networks allow a variety of input traffic in addition to the Poisson-distributed traffic. In this thesis, individual user traffic process is modelled as a two-state Markov modulated Poisson process; the aggregate user traffic process is modeled as a batch Bernoulli renewal process under short-term condition and as a fluid process under long-term heavy traffic condition. The signalling traffic at the UNI carries call control messages and network management messages originated from the user nodes. The signalling traffic must be serviced quickly since they directly affect call establishment and network efficiency. Up to now, all related congestion control researches only consider user traffic. Consequently, the primary objective for this thesis is to study the effect of the higher-priority signalling traffic on the multiplexing of user traffic at the UNI. A novel modeling of user traffic multiplexing through the ATM statistical multiplexer at the UNI is proposed: it is characterized by a queueing model with random service disruptions due to the transport of higher priority signalling traffic. The congestion performance requirements of the user traffic for the UNI are studied in terms of the stochastic cell loss requirement and the deterministic upper-bound cell delay requirement. However, in order to investigate the stochastic cell loss phenomenon due to buffer overflow, the stochastic queue behaviour must first be examined. Consequently, a novel algorithm to solve the stationary distribution of the queue length process under short-term heavy traffic and finite buffer capacity conditions is presented. A novel UNI call admission control scheme is proposed, and its objective is to maintain the required network performance assigned to the UNI access-node by exerting call admission control in the call access phase of each user traffic source. It is analyzed using an input-limit static control model employing stochastic ordering between the cell loss ratio random variable and the desired threshold random variable as a criterion to decide if a new call should be admitted. The cell loss ratio random variable has been chosen as the performance objective rather than the long-term-time-averaged cell loss ratio, so as to take into account of the dynamic nature of bursty traffic sources. A novel UNI intra-node buffer control scheme is proposed, and its objective is to optimize the network performance of the UNI access-node by exerting buffer control in the aggregate information transfer phase of the user traffic sources. It is analyzed by means of a sequential decision process model characterized by a stationary, Markovian and deterministic threshold control policy.

Text

2010-11-24

For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

http://hdl.handle.net/2429/30126

Electrical and Computer Engineering

University of British Columbia

Graduate