Low Latency is gaining importance amongst operators, and they are focused on reducing latency in each of part of the network including the Wi-Fi links in the home, DOCSIS links in the access network and core network segments. Providing lower latency and hence measuring the latency in each portion of network is a vital requirement for MSOs. Operators will need to troubleshoot latency issues and need the ability to identify latency within their networks vs. outside of their networks.
This paper aims to share the experience from developing a simple end-to-end latency measurement framework. A new measurement protocol defined in the IETF is STAMP (Simple Two-Way Active Measurement Protocol, RFC 8762). The paper will provide the lessons learnt from developing a proof of concept for latency measurement using STAMP. It will describe the high-level measurement architecture and locations for measurement agents and peers. A STAMP-reflector could be implemented in a gateway or a device behind it and a STAMP sender can be implemented somewhere in the network (e.g., in a hub, north of a CMTS). An operator can start with small number of measurement entities and scale up as needed. If a Session reflector can by dynamically instantiated in a gateway, then one can run measurements on-demand. The paper will also investigate methods to kick off different latency tests and have the measurement end points report latency data. It will also look into how latency data from various sources can be aggregated and reported. It will also discuss measurement control and reporting methods based on LMAP (Large-Scale Measurement of Broadband Performance). The paper will provide an understanding of MSO needs around latency measurement, an overview of the most appropriate metrics to report, and how to deploy measurement technologies to meet those needs. The paper also reports on a prototype STAMP measurement system which is deployed and collecting latency data.