Measuring the Arrival Quality of Real-time Packet Trains – a Global PerspectiveDate: 2013-11-19
Time: 1:00 pm
Location: Holmes Hall 389
Speaker: Professor Ulrich Speidel, Department of Computer Science, University of Auckland
Real-time applications on the Internet include everyday applications such as Voice over IP telephony, but also more advanced technologies such as remote manipulation, e.g., for remote surgery. These applications work best if their packet trains arrive with minimum latency, low packet loss, constant inter-arrival times and all packets in the order in which they were transmitted. Latency is largely unavoidable due to the physical distance, but not meeting the remaining requirements perfectly requires applications to buffer packets until sufficient data for meaningful processing (e.g., audio playback) has accumulated at the receiver.
Conventional "improvements" to Internet infrastructure, such as the addition of new links and load balancing can be a double-edged sword: While they create extra bandwidth and reduce congestion and sometimes latency, they also create additional router queues and alternative paths, potentially affecting inter-arrival times and in-order delivery. This is in particular a problem if the destination is served by long thin networks. Our project is a longitudinal study that attempts to track the long-term global trend in the arrival quality of real-time long distance packet streams.