ACTS Project AC 085 Wireless ATM Network

The purpose of the WAND demonstrator performance measurements was to evaluate how well the requirements of the WAND system design are met. This deliverable sums up the results of the performance evaluation. A special focus is put on evaluation of the WAND specific components that were designed and implemented within the project. The performance of the WAND radio sub-system, consisting of WAND radio Modem and MASCARA medium access scheme, and mobility enhanced ATM signalling stack is evaluated and the results are introduced. The end-toend performance of the WAND system was evaluated with a network benchmark tool called Netperf and the throughput and transfer delay results are presented. Keyword list: Wireless ATM, Performance Evaluation CEC Deliverable No: AC085/NMP/WDA/DS/P/036/b1 Title of Deliverable: 5D2 WAND Demonstrator Performance Report AC085/NMP/DS/P/190/1.0 2 Part II (Executive Summary) The goal of the ACTS Magic WAND project was to specify and implement a high speed indoor wireless ATM access network with QoS support for mobile multimedia services, such as video conferencing. The WAND design work covers the entire network ranging from the radio link specification to mobility enhanced core network functions. Expanding wired ATM into the wireless world is a real challenge for the WAND system performance, since ATM technique has been traditionally used in a low bit error rate conditions. Also, the applications set high requirements for the performance. Real–time applications, such as voice and video, require a low delay and delay jitter, while data applications like WWW or file transfer need to have relatively low BER. This deliverable sums up the results of the WAND demonstrator performance evaluation. A special focus is put on the evaluation of the WAND specific components designed and implemented within the project. The main system components are WAND radio sub-system, consisting of WAND Modem and MASCARA medium access scheme, and mobility enhanced ATM signalling stack. The performance of each WAND specific system component was evaluated. The WAND Modem evaluation showed up that the performance of all the modules of the WAND Modem individually worked up to specification. Current implementation of the Modem meets the minimum criteria for user trials and system testing; a raw Packet Error Rate (PER) of 0 % was achieved under good propagation conditions and less than 1 % PER is possible in many cases. In some cases (deep fades) performance was found to be much lower and the PER goes up to the 10% range. It was also found that although all modules of the Modem individually worked up to specification, small imperfections in each module added up and caused problems in the link budget. The results of the MASCARA performance evaluation indicated maximum throughput of 10 Mbit/s in the measurement conditions. The execution time of the association procedure, one of the heaviest operation from the MASCARA point of view, was completed in 8.4 μs as the requirement for the operation is 88 μs. The performance of the mobility enhanced ATM signalling stack was evaluated by measuring mobile terminal registration delay, connection setup times and delays caused by handover procedure. The results show that the control plane implementation, although not being the most optimal one, provides a good basis for a WAND demonstrator. The end-to-end performance of the WAND system was evaluated with a network benchmark tool called Netperf. The primary focus of the tests was to measure throughput and end-to-end transfer delay with TCP and UDP protocols used on top of WAND system. Two equipment setups were included in the measurements; in the first case two workstations were connected via fixed link using mobility-enhanced ATM switch, the Network Server, and the User Mobile Terminals. In the second case, the MASCARA subsystem, Access Point and MASCARA mobile terminal, were also included. The throughput in the first case was on average 30 Mbit/s with AAL5 and TCP and up to 50 Mbit/s with UDP, which are comparable to commercial ATM CEC Deliverable No: AC085/NMP/WDA/DS/P/036/b1 Title of Deliverable: 5D2 WAND Demonstrator Performance Report AC085/NMP/DS/P/190/1.0 3 systems. The mean end-to-end transfer delay was 0.7 ms when TCP was used. With the MASCARA subsystem, the throughput performance was adequate; average transfer rates stayed in 12...15 Mbit/s with TCP, UDP and AAL5 and the additional delay caused for TCP by the MASCARA was 2.6 ms. It was also found that using TCP protocol in a fairly unreliable conditions as Wireless ATM system, can lead to a reduced performance due to the sliding window mechanism used in TCP protocol. CEC Deliverable No: AC085/NMP/WDA/DS/P/036/b1 Title of Deliverable: 5D2 WAND Demonstrator Performance Report AC085/NMP/DS/P/190/1.0 4 Part III (Full Description of deliverable content) TABLE OF CONTENTS 1. OUTLINE OF WAND PERFORMANCE EVALUATION.............................6 1.1 WAND System Overview................................................................................6 1.2 Demonstrator System Architecture.................................................................7 1.3 WAND Protocol Stack ...................................................................................8 2. WAND RADIO SUB-SYSTEM PERFORMANCE..........................................9 2.1 WAND Modem performance..........................................................................9 2.1.