Variable Packet Size Sample Clauses

Variable Packet Size. ‌ The Variable Packet Size (VPS) technique aims at measuring the capacity of every link of the path. It was first described by [5] and used in tools like pathchar [31], − clink [16] and pchar [39]. The basic idea in VPS is to measure the delay variation when increasing the size of the packets: intuitively, if the capacity of a link is C, the transmission delay increases linearly with the size of the packet as 1/C. It is thus possible to compute the capacity by sending packets of different sizes and observing the change in delay. In a path composed of multiple links, the capacity Ci of link i can be measured by estimating first the capacity of all previous links and then by subtracting from the delay variation measured up to link i, the delay imposed by all previous links up to link i 1. Similarly to the traceroute tool, in VPS tools the TTL field in the IP header is adjusted so to measure the RTT up to the desired hop. Then the size of the packets is modified and the RTT is measured again. If there is no queuing due to cross-traffic, the difference in delay is due solely to the difference in size and the capacity of the link is obtained by recursively measuring all previous links. VPS can thus suffer from error propagation. Additionally, in order to obtain delay samples unaffected by cross-traffic, several hundreds (or thousands [16]) packets must be sent and only the sample with the minimum delay is considered. Finally, as shown in [47], VPS is inaccurate in presence of layer-2 store-and-forward devices, which are invisible at layer-3 but add serialization delay.