An Investigation on Core Network Latency

The growing adoption of 5G and cloud services places an increasing importance on the attainable point-to-point latency in Internet service provider networks. It directly impacts whether latency-critical services can be offered as well as the number and location of the corresponding data centers. In order to investigate the limits and variability of latency values in an actual core network, we collected round-trip time values published by a large North American Internet service provider spanning the duration of more than one year. We present a statistical analysis of this data set from which we infer a potential fiber topology. We use this topology to hypothesize on the efficacy of different means of reducing latency and determine their effect on the viability of low-latency services.