Thomas Theimer

Performance analysis of buffered Banyan networks

The authors give a classification of the most common Banyan networks and outline an analysis approach for the rather general class of delta-b networks with multiple buffers. On the basis of this approach, they discuss the effects of the approximation involved and present a refined analysis algorithm for the special case of a single buffered delta-2 network. The results of both algorithms are compared with simulation results to assess their accuracy. The comparison shows clearly that the accuracy of the analysis is mainly determined by the modeling of the dependencies between consecutive clock cycles. >

A review of voice, data and video traffic models for ATM

This paper describes the main statistical features of voice, data and video sources, defines a set of selection criteria for selecting source models for these sources, describes a number of relevant source models for representing these sources and recommends the best source models based on the results of the parameterisation of these models from measured data.

Approximate analysis of the end-to-end delay in ATM networks

A very general approach for an approximate analysis of end-to-end delay jitter in asynchronous transfer mode (ATM) networks for bursty input traffic is presented. Since the cell loss requirements within ATM networks are very restrictive, interest is in the tail of the complementary probability distribution function of the end-to-end delay to estimate the cell loss probability at the reassembly buffer. An analytical approach for an isolated network node as well as for a whole reference connection is presented. The approximation results are in good agreement with simulation results and indicate that the end-to-end delay jitter is in the range of the maximum transfer delay of one queuing node, if the bursty input traffic is controlled by adequate traffic control mechanisms.<<ETX>>

Buffering concepts for ATM switching networks

The authors study various buffering strategies for switch fabrics of the Banyan type, where the buffers are located in the individual switching elements. The influence of the different strategies on the performance of the network has been evaluated under equivalent conditions to allow a comparison of the results, and the packet loss probability for networks constructed of switching elements with input buffers has been analyzed. The investigations indicate that significant performance improvements can be achieved by applying advanced buffering concepts. The choice of the best buffering strategy depends on the size of the switching elements as well as on economic and technological constraints.<<ETX>>

Bandwidth allocation mechanism for LAN/MAN interworking with an ATM network

Abstract In the evolutionary path to Broadband-ISDN based on the Asynchronous Transfer Mode, the interconnection of LANs and MANs will be one of the first high-speed applications. However, interworking between LANs/MANs and ATM is an open problem, since LANs and MANs use a connectionless transfer mode, whereas ATM is a connection-oriented technique. This paper focuses on a main problem arising within the interworking unit connecting a LAN/MAN and the ATM network: the bandwidth allocation for the ATM virtual channel established between the IWU and a CL-server. A possible bandwidth allocation mechanism based on an estimation of the bandwidth requirements of the incoming traffic stream which is routed to the CL-server is described. To investigate the performance behaviour of this mechanism, an analytical model based on a fluid flow approach is developed which will be validated by simulation. Comparing different bandwidth allocation strategies we show, besides other results, the trade-off between buffer size and the allocated bandwidth on the ATM virtual channel to provide a certain quality of service in terms of packet loss and delay.

Switching block studies, network performance evaluation and traffic engineering for ATM

This paper summarizes the work done in RACE project 1022 “Technology for ATD” on performance studies of ATM switching block architectures and the studies on network performance within a small ATM network, the so called RACE ATD Technology Testbed (RATT). The aim of the studies was to investigate the behaviour of a typical ATM network under nominal load. By using the tromboning links in the RATT we were also able to study long distance connections even in the limited size of the RATT network. Special attention has been given to define a traffic scenario which can also be set up for the planned experiments. The goal of the planned experiments is to validate the performance results predicted in our studies.