On lower estimating internet queuing delay

With the advent of online social media, latency became a primary factor and large Content Delivery Network (CDN) providers optimize their networks for. While many ingredients (e.g. propagation delay, transmission delay, processing delay) of the end-to-end latency can be adequately characterized, capturing queuing delay seems to be a difficult task due to the complex and often unpredictable nature of Internet paths and the imperfect tools used for measurements. Dealing with the queuing delay is so challenging that most papers concerning end-to-end latency completely ignore it for “simplicity”. In this paper we take this view to the extreme and try to interpret the cause of various delays between the same endpoints as measurement artifacts, such as continuous changes in end-to-end paths and the imperfection of measurement tools. We arrive to the conclusion that there is a significant amount of points in end-to-end delay measurements, which cannot be explained by these artifacts even if we do our best for doing so. The only plausible explanation for these points is the presence of significant queuing delay comparable with the delay caused by all the other factors (propagation, transmission and processing).