Latency Modeling for Mobile Edge Computing Using LTE Measurements

A promise of mobile edge computing (MEC) is improved communication latency. In particular, vehicular safety applications have strict latency requirements. Despite the improvements in latency by MEC, the latency will vary based on traffic load and signal conditions. We seek models that will help us answer the question: is the latency environment acceptable for our application? To this end, we collected 619 hours of pings over an LTE network and we evaluated 13 possible models. Our results show that when modeling the whole distribution, a mixture model of LogNormal distributions works best. When the emphasis is on the tail of the distribution, the double Pareto LogNormal distribution should be used. Finally, when only the tail needs to be modeled, then the generalized Pareto distribution should be used.