A Closer Look: Service Assurance for Virtualized Networks

Posted by Mae Kowalke on Thursday, September 08, 2016 with No comments

So far, 2016 is proving to be a transitional year for mobile network operators and their use of performance assurance tools, as they make the move toward 5G and virtualization through C-RAN, virtualized customer premises equipment (vCPE), and other software-focused solutions. That was the overall conclusion reached during a recent conversation between Scott Sumner, VP of Strategic Marketing at Accedian, and Dan Meyer, Editor-in-Chief of Telecom Software, Policy, and Wireless Carriers at RCRWireless News. 

As they begin rolling out services like voice over LTE (VoLTE), Sumner noted, quality of experience (QoE) is proving to be more critical than ever, and also more challenging to achieveat least using conventional tools and methodologies. 

“QoE is key as we get into virtualization and networks become more granularly controlled at application and policy levels,” Sumner explained. “Operators need to make sure the people who are paying for service are getting what they need, and also that all the diverse applications are coexisting seamlessly.”

VoLTE is a good example of this effect, Sumner said, because many roll-outs are based on highly virtual platforms. Operators tend to expect that, in such situations, networks and services will behave as they did in the static world of hardware-based networks. But, they’re finding that to not be true.

“Not only is the virtualization layer itself hard to manage, but also it interacts on new levels almost like a living, breathing entity,” Sumner explained. “All these different network functions talk to each other, and to some degree they are self organizing and autonomous. This creates a lot of micro-fractures or micro-issues that together can cascade with a knock-on effect and actually take down a network or service at large scale.”

Operators have been caught off-guard by these issues because the measurements and metrics used for traditional packet-based transport, backhaul assurance, or voice services, as just not enough to see what’s happening in these newer, more dynamic networks. What’s needed are three things: 1) a wider range of performance indicators, 2) much more granular metrics resolution, and 3) a statistical reporting-based approach to identifying root causes of service issues or failures.

In one example cited by Sumner, a large Asian operator was seeing recurring VoLTE failures on its very modern network. Accedian’s performance monitoring and assurance technology, which the operator used to instrument its network, provides more than 200 statistical viewpoints and key performance indicators (KPIs). In this case, the only metric that pulled up a correlation with the failures was a ‘lost burst’ measurement: essentially, a group of packets lost in a row, sequentially. Even though the amount of packet loss seemed insignificant, the fact that it happened instantaneously caused VoLTE calls to drop.

The minute nature of this issue illustrates why traditional performance monitoring methods are mismatched for next generation, dynamic, virtualized networks and services.

“Thresholds can’t be applied like they used to,” Sumner pointed out. “You can’t catch microburst loss with simple thresholds that say, for example, at 1 percent packet loss, the system should set off an alarm. What’s needed now is the ability to go into the whole data lake and see how data records correlate with call failures, packet loss, router firmware, and other factors. Machine learning and statistics are becoming the future of running these networks successfully.”

Sumner’s advice to operators: instrument the network with modern tools before rolling out services like VoLTE.

“Don’t work in the dark,” he stressed. “Turn on the lights. See what’s going on. It’s fast, easy, and cheap. Accedian has instrumented networks with hundreds of thousands of base stations
some of the largest networks in the worldin a period of 8 weeks at a cost that’s significantly lower than anything the operator has used before for monitoring. They can roll this stuff out quickly on their own cloud, and start seeing things right away.” 

Watch the full video for more discussion about virtualization and network performance assurance, including the results Accedian has achieved by helping operators share their virtualization learnings through a center of excellence, and why operators and vendors to work together on interoperable virtualization solutions.