A Light-Weight Statistical Latency Measurement Platform at Scale

The statistical value of latencies between two sets hosts over a given period, which is referred as to the latency, can benefit many applications in next-generation networks, for example, Network-in-a-Box-based resource provisioning. However, existing methods hardly achieve low measurement cost and high prediction accuracy simultaneously large-scale scenarios. In this article, we design light-weight latency platform named DMS (DNS-based Measurement at Scale). achieves by introducing metric space select closest open recursive DNS (Domain Name System) server host, predicting end-to-end via measured corresponding servers. To reduce overall overhead, clusters with infrastructure network cluster center, thus achieving good scalability large scale simultaneously. evaluate performance DMS, implement prototype system network. Compared widely adopted method King, relative error 18.5% real-time 33% prediction.