A Survey of Round Trip Time Prediction Systems

Abstract

The performance of many networked applications improves when Round Trip Time (RTT) is reduced. With lower RTTs, human-to-human interactions become more realistic, query-response activities (such as DNS lookups) are more responsive, and transport protocols (such as transmission control protocol) can better utilise a path’s available capacity for long-lived file transfers. Additionally, RTT prediction systems are beneficial in distributed networks in order to improve performance and resource allocation. RTT prediction systems predict RTT for all nodes in the network without additional exchange of packets. We discuss 13 RTT prediction systems under three classes: 1) localisation measurement systems; 2) network coordinate systems; and 3) matrix factorisation systems. RTT prediction systems have been applied with positive results to various applications and services, such as peer-to-peer networks (e.g., BitTorrent clients, content distribution networks, and Internet of Things) or client-server networks (e.g., game server discovery).