Anomalies due to delay and loss in AAL2 packet voice systems:performance models and methods of mitigation

摘要

This paper investigates the packet delay variation and packet lossnprobabilities to be expected in typical applications of voice trunkingnusing ATM adaptation layer type 2 (AAL2) and the consequences these maynhave, such as the need for sequence numbers in voice packet headers.nSimulations have been run for the case of 32-kb/s ADPCM encoded voicenacross a physical link whose bandwidth is 1/4, 1/2, or 1 times DS1n(1.536 Mb/s). The results show the number of calls that can be admittednwhile meeting the ninety-ninth percentile queuing delay objective of 20nms or less. The delay budget that can be allocated to packet queuingndelay at access is about 20 ms, given a one-way end-to-end delay budgetnof 100 to 150 ms for conversational voice. The number of calls is alsonlimited so that the ratio of late or lost packets does not exceedn10-3. Voice quality is known to be tolerant to packet lossesnof up to one in 100. When the packet delay variation is significant, asnit threatens to be in this situation, sequence numbers can help anreceiver to detect and to recover from anomalies-lost, early, or latenpackets-and make fewer reconstruction errors than if sequence numbersnwere absent or ignored. Three circumstances are identified where packetnslips can occur with ripple effects. An analysis of these cases leads tonlower bounds on the sequence number modulus. The appropriate value ofnthe sequence number modulus depends on the desired call load andnperformance objectives, and can be deduced from the results presented