Anssi Tauriainen is director of the Analytics and Advanced Services Business Unit at EXFO. Here, he tells VanillaPlus that analytics systems that can automatically derive insights across multiple network domains are vital for enabling communications providers to have visibility into their operations across physical, virtual and hybrid networks
VanillaPlus: What role do you see analytics capabilities playing in enabling NFV?
Anssi Tauriainen: In our vision, analytics is going to play several key roles. There is the more traditional and well-established role of providing visibility into the Service Experience (SX). This becomes more complex as NFV rolls out because that visibility now has to be collected from physical, hybrid and virtualised networks. Today, we’re able to collect the needed information from all of these environments and create the closed loop analytics case and provide that visibility to service orchestrator, which in turn will use the information to configure and optimise the network. EXFO is differentiated because we have complete end-to-end view across all segments of networks, we are the only ones who have the visibility through user, signalling, IP and physical layers of network, and our analytics contains network topology and service modeling, which means we can analyse service, subscriber and network experience at the same time. We call this 3D Analytics.
We have components in our analytics platform that have a constant discussion with service orchestration so whenever new virtual network functions (VNFs) are deployed we can dynamically orchestrate tests, monitoring and SX management criteria for newly created instances. We have a portfolio of virtual verifiers and passive monitoring equipment so, depending on what is being done, we can provision the right testing and monitoring scenario for the situation.
VP: How important is the capability to be able to test and verify across the different types of network – physical, virtual and hybrid?
AT: It’s a mandatory requirement because there are not likely to be a large number of greenfield virtualised networks. All will be hybrid, whether that’s for ten, 20 or 30 years is a question, but it’s certain that fully virtualised networks won’t happen overnight.
VP: How can you measure service experience? Please can you give examples of indicators used to generate a Service Quality Index?
AT: We want to create the end-to-end view so instead of focusing on a specific subnetwork or part of the network, we want to cover the whole chain from one device to another across the different network layers: the user data, the signalling data, the IP traffic and the physical layer. That’s the starting point from which we’re building the end-to-end service experience.
We have a hierarchy for this called the Service Experience Index (SxI) which extends across the layers and is conducted from Key Quality Indicators (KQI), Key Performance Indicators (KPIs) and raw performance counters.
Instead of looking at values in real-time alone, we also have historical numbers available so CSPs can gauge also past performance, deviation and trends. This gives us the capability to put changes in experience into context. Just looking at today’s data isn’t enough. At the highest level, we look at for example availability, reliability and performance from service, subscriber and network points of view simultaneously.
As an example, with a data service we look at service availability as one of the service experience indicators based on accessibility and set-up time of the connection.
VP: How can you tell me what customers will be impacted by a performance degradation on a device?
AT: That is an easy question to answer. Every single call detail record (CDR) or xDR contains a subscriber ID. When you look at an individual service usage session you combine the xDRs together and that tells you what’s being experienced.
A more challenging question to answer is what you’re going to do about discovered findings since there are likely tens of thousands of sessions with degraded service experience ongoing in the network at any given time. The situation becomes all the more challenging as new development such as Internet of Things (IoT) add millions of devices to the network. No one will have the manpower to analyse all of the data at the same time.
Therefore we come back to segmentation to make the data meaningful to CSPs. For example, we can look for prioritisation criteria such as customer or service business value or common denominators such as combination of application, device or firmware versions, or network subsegment or technology that may reveal the root cause of an issue.
There are endless number of variations in this equation and therefore traditional heuristic approach will not be sufficient to discover the denominators. More advanced analytics methods such as machine learning and pattern recognition will be needed.
VP: How can your product reduce Mean Time to Repair?
AT: As mentioned earlier, we’re the only player who can combine analytics of these different network layers – the mobility, IP transport and physical fibre networks. That ability to collect the end-to-end data from the different layers combined with advanced analytics methods, such as automated root cause analysis means that we can also make sense out of an issue much faster than partial or manual analytics.
For example, if a fibre link between the UK and Germany is cut, it’s a crisis for a CSP. They know immediately about it but they can’t tell which subscribers and services are impacted and if there is a way to mitigate the impact. In addition to user, signalling and IP traffic, our solution contains the network topology so we can see e.g. the subscribers who just lost their single connection and take this information to SOC for further actions.
They can then notify their customers that that they’re working on re-route or replacement connection.
VP: How do you integrate your product within a hybrid network composed of physical and virtual assets?
AT: In addition to analytics and EXFO Mobile Agent, our Service Assurance portfolio contains also passive monitoring products and active verifiers. Today, both are tapped into the physical network using physical interfaces but the approach will be the same for virtual networks. We’ll simply tap in using virtual interfaces. The internal function remains the same – the analogy is very similar but the execution is different.
VP: What’s the significance of active verifiers in contrast to passive monitors?
AT: Active verifiers provide very scalable and cost effective tools for ensuring service level agreements are met and various network functions are available and performing well. Passive monitoring of all connections that go through the network provides a lot of information but is also very intensive. Even in the cloud it takes huge processing power to manage that and processing power won’t be free, even in the virtual world.
Active verifiers provide a very cost effective means of ensuring the service experience is being provided to the expected levels. That’s why we’ve added them to our portfolio, which we believe offers the broadest range and scope for CSPs as they continue on their transformation journeys and embrace new technologies, new business models and new types of services.
