An Asian operator with a 100,000 LTE base station network is carrying out live testing and monitoring of its network using virtualised instrumentation.
The operator, who TMN is unable to name, is using virtual probes from Accedian to send test packets into a network that has more than a quarter of a million endpoints, generating 20 billion metrics per day to feed into its big data analytics systems. Metrics being monitored include delay variation, packet loss, MoS scores. The operator is using the feeds for ongoing analysis, and to assure backhaul links when new parts of he network are brought online. Accedian has claimed the deployment is the first such fully virtualised assurance solution in a live commercial network.
Some service measurements – such as on VoLTE – are made 50-60 times a second, with KPI reports every minute, according to Henrik Nydell, Senior Product Manager at Accedian Networks. A single virtual probe instance can instrument around 4,000 base station connections, Nydell said, by being plugged into the network at the aggregation layer. Accedian’s Skylight Director then orchestrates those probes, setting up automated measurements on the network. If the virtual instance is sited deeper in the network, at a central location, then Small Form factor Pluggable (SFP) agents provide the feed from the network elements to the centralised cloud instance.
The deployment shows the ability of virtual test agents to scale and flex to service providers’ needs, Nydell said. However, he added that new measurement protocols will be required in virtualised infrastructures expected to underpin 5G and mass IoT networks. IETF test standards for networks – such as Two-Way Active Measurement Protocol (TWAMP) RFC5357 – will need to be upgraded to ensure that standards that work for 4G work well in 5G.
“For 5G virtualisation will be key, but you can’t really put a probe in the cloud. 5G will be a living entity. Not only users will move around but services will too as operators spin up new servers and instances dynamically. Things will be floating around and you cannot see that at the network level, so that instrumentation capability will need to be built-in. For IoT the responsibility for monitoring the KPIs will need to be be built inside the app itself.”
Nydell added that one example where more work is required is time stamping in virtual architectures. He said that Accedian is amongst those working with Linux groups to find a way to provide accurate time sampling for server hardware supporting virtualised architectures.