The importance of being really random: methodological aspects of IP-layer 2G and 3G network delay assessment

The accurateness of round-trip- and one way delay measurements for 2G and 3G networks rely to a much larger extent on employing a sound methodological framework than this is the case for other types of networks. Typical mobile access networks differ significantly from core networks, most prominently with respect to delay. In this paper we present payload-dependent delay measurement results for public 2G and 3G networks which illustrate that accurate IP-layer delay measurements in mobile networks must use high sample counts and randomness in start times for uplink and for downlink. Most notably this concerns ICMP round-trip delay measurements which, due to the synchronization of ICMP requests with the network clock when leaving the mobile network's uplink, fail to meet the random start time criterion for ICMP replies in the downlink (or vice-versa). This synchronization effect leads to significant clustering of one-way delay values for the reply leg and causes minor delay differences in the core network to have potential significant impact on ICMP round-trip delays. Therefore, highly accurate simulations and emulations must model uplink and downlink of time-slotted networks as two interrelated links based on a common timebase.