Media splashes by Dish and Rakuten may be proof that open-source software has revolutionized the telecom industry, but most brownfield operators have yet to see meaningful cost reductions related to free, community-developed software.
However, instead of despairing at spiraling technology costs of the 5G standalone (SA) core, operators should consider the results of a series of tests by Telenor Group. Telenor was Norway’s incumbent operator and now has properties across both the Nordics and Asia. Its R&D department has just bagged a Mobile World Congress 2021 Glomo Award for ‘Best Network Software Breakthrough.’
Earlier this year, the Telenor team built what it believes is the world’s most multi-vendor 5G SA set-up, involving RedHat, Casa, Kaloom, Oracle, Huawei, Enea and others. The deployment used many open-source software components and most importantly, it was in a cloud-native architecture. (“Doing a 5G core without going cloud native should be considered a crime,” deadpans Pål Grønsund, senior research scientist at Telenor.)
Telenor’s tests demonstrated that it is feasible to deploy a service-based architecture in a vendor-agnostic environment for some pretty cool use-cases such as network slicing, IMSI concealment and protection against signaling storms.
Fierce recently caught up with Kashif Mahmood, Telenor Group’s senior research scientist, in order to better understand the implications for operators currently building out 5G. Here’s an extract of the discussion.
Fierce Wireless (FW): Traditional 5G core vendors have often pushed their own closed container solutions touting the advantages of interworking and performance. Was it difficult to convince the companies involved to commit to a vendor-neutral environment?
Kashif Mahmood (KM): It’s true that in the past, vendors have talked about the responsibility matrix or the performance aspects to push their one-stop solution. After the initial resistance though, the vendors we worked with understood our objective and stepped up to the challenge. Our results demonstrated how easy it was to onboard the functions into the OpenShift environment, in fact we were surprised at how transparent it was.
FW: You could have opted for a generic Kubernetes platform, but instead you chose OpenShift, which is a curated version from Red Hat. What would be your recommendation to other operators choosing a container platform?
KM: Although it’s theoretically possible to use open source Kubernetes directly, as an operator we would always recommend using a version that’s fully supported. It may contribute to some sort of vendor lock-in, but in the end, even if it’s open source we still need a third party to own it. Red Hat really provided value in pulling together the vendor agnostic environment.
FW: You did another 5G core SA test in 2020. What are some of the biggest changes you’ve seen in a year?
KM: We have seen a big improvement in the maturity and the diversity of the solutions. Last year’s trial involved a Kubernetes environment provided by the software vendor. At the time, many people had doubts about open-source PaaS (platform-as-a-service) components such as Grafana, Jaeger and Prometheus. For this latest, award-winning proof of concept, we used all these components, and the CNFs (containerized network functions) were onboarded on a vendor-neutral Kubernetes platform powered by OpenShift from Red Hat.
FW: What were some of the biggest challenges you faced with these tests?
KM: Given the high degree of vendor interworking, I was pleased that we did not encounter any insurmountable problems. Having a highly motivated team played a big part in this success. We did face some interworking issues; the software components did not always work straight out of the box. Some CNFs had issues registering with the NRF (The Network Repository Function, which is an important component of the service-based 5G core architecture.) These CNFs had to be manually registered with the NRF.
Also, network slicing which is a key feature of 5G is not well-supported on the end device side.
There were other minor challenges, but what struck us was how quickly the vendors were able to address the problems. One day, I spoke to Casa systems about an issue we were having with the SMF (Session Management Function). They had their U.S. engineering team work on it during our nighttime, and by the next day we had new software available that fixed the problem. It really demonstrated the kind of agility we can expect in our future deployments.
FW: That’s pretty impressive! Do you have any other recommendations for operators focused on designing and building their 5G core?
KM: Virtualization has helped us move away from hardware silos, but we need to watch out for software silos. I would recommend that operators take the bold step of building a best-of-breed 5G core on a vendor-neutral container orchestration platform.
It is also important that automation be there on day one to get rid of manual, repetitive tasks – approaching the zero-touch network. Ansible, another open-source software, will play a critical role in this regard. Due to limited time, we only managed to integrate some of the components with Ansible, but we could really see the importance of making sure the playbooks are ready from the first day of deployment.
The message from Telenor is that operators should take heart, especially considering Kashif’s last two points.
Open-source software has allowed the telecom industry to leap forward in developing a new, services-based architecture. A micro-services approach means we can expect an agility we’ve never had. Fully automated deployments will dramatically decrease the wait for network changes and new service introduction. The network may cost more to acquire and integrate, but innovations should arrive with unprecedented speed.
And as we all know, time is money.