Published on June 16, 2017
How do you upgrade your network for one-way delay metrics without breaking the bank?
Today’s services, and consequently the networks that carry them, are highly asymmetrical. Most bandwidth-hungry applications, such as IP Video, require much more bandwidth downstream than upstream. At the same time, self-optimization technologies, such as software defined networking (SDN), no longer require the upstream and downstream traffic to follow the same path through the network. The result of these two effects is that service path latency is now asymmetrical as well.
This asymmetry can have a profound impact on today’s networks and services. 5G networks rely heavily on sub-millisecond, one-way latency to maintain service quality. Unexpected increases in one-way latency can result in significant degradation in throughput or dropped calls. At the same time, emerging applications, such as self-driving cars, also demand very low one-way latency to ensure that safety-related issues are communicated in time to allow the car to react.
Round-trip latency metrics only provide half the information needed to manage today’s CSP networks (literally). In the past, it was acceptable to estimate one-way latency from two-way latency tests simply by dividing the two-way result in half. While this was acceptable for services that did not have today’s stringent requirements for one-way latency and were carried in symmetrical, highly engineered networks, this technique does not work for today’s services and networks.
Anyone familiar with latency measurements knows that one-way latency requires synchronized timestamps at each end to determine the path latency. Up until now, this synchronization could only be achieved with expensive GPS solutions, which are susceptible to satellite visibility in urban locations or malicious jamming of the satellite signal, or packet-based solutions, such as IEEE 1588v2 PTP, which can be expensive and complex to implement, especially as a retrofit to existing networks.
To address this potentially costly and complex dilemma, EXFO has introduced a solution for their Active Verify platforms, which can deliver one-way performance metrics using standard two-way testing protocols, such as ITU-T Y.1731 or IETF RFC 5357 Two Way Active Measurement Protocol (TWAMP), and without the need for a synchronized timestamp at the far end. In fact, the remote site can be any industry standard reflector for Y.1731 or TWAMP. The EXFO Universal Virtual Synchronization Feature can achieve derived one-way latency metrics within 200us of actual performance using free running clocks on both ends.
To find out more about EXFO’s Universal Virtual Synchronization feature, visit our website or download the white paper: Measuring one-way delay for optimal delivery of revenue-generating services.