Skip to main content
Published on

Network-wide latency maps for mesh, hybrid and cloud networks

A key driver behind mesh, hybrid and cloud based networks is the ability to optimize service paths and application server locations to ensure performance metrics continue to be met. Knowing in advance how moving an application server or changing the service path will impact latency is critical to ensuring that the customer’s overall quality of experience (QoE) is maintained. However, if the goal is to be able to make these optimization changes in real time, before the customer even notices a service degradation, then having a continuous view of per path latency is essential.

Compounding this problem is the fact that services today are becoming highly asymmetrical with respect to bandwidth and, consequently, latency. At the same time, optimized service path routing means the upstream and downstream paths likely do not follow the same route, again leading to asymmetry in upstream and downstream latency.

To be able to build an accurate map of the per link, per direction latency for a network requires the ability to derive 1-way latency metrics from your service OAM solution. Traditional wisdom tells us that to measure 1-way latency, you need highly synchronized timestamps at both ends of the service path. Unfortunately, many networks today do not have this capability. Instead, they rely on 2-way, or roundtrip, latency measurements, divided by 2 to estimate 1-way latency. Given the highly asymmetric nature of service latency and service path latency, this technique can lead to incorrect assumptions and unforeseen service quality issues.

So how does a carrier go from 2-way latency measurements to accurate 1-way latency measurements? Again, conventional wisdom tells us that we need to instrument every measurement site with a synchronized timestamp solution. Typically, this would be either a GPS receiver providing synchronization or a packet timing protocol-(based synchronization solution, such as IEEE 1588v2, both of which can be expensive and complex to implement.

Fortunately, this is not the only solution. EXFO’s Universal Virtual Sync solution, available on many of their BV series of Active Verifiers, can derive highly accurate 1-way latency measurements using industry standard 2-way latency tests, such as ITU-T Y.1731 or RFC 5357 TWAMP (Two-Way Active Measurement Protocol); and, the far end reflector can be any device that supports these protocols. No additional capabilities required.

In side by side testing of latency measurements with one setup using GPS synchronized clocks at both ends and the other using free running clocks with the EXFO Universal Virtual Sync feature on the generator end, 1-way latency measurements from both solutions were shown to be within 200 us of each other.

If you would like to learn more about how the EXFO Universal Virtual Sync feature can provide you with greater insight into your network’s performance without breaking the bank, download: Measuring one-way delay for optimal delivery of revenue-generating services or visit us at