As an industry, we’re tumbling headfirst into 5G and edge without necessarily understanding all that is ahead of us. Running proofs-of-concept and upgrading radios to get faster speeds on wireless mobile plans is one thing, understanding infrastructure upgrades needed to make 5G a profitable business case is another.
I agree with Chetan Sharma’s article that telcos who believe 5G is a “build it and they will come” bet will fail. And so, infrastructure initiatives that provide flexible and streamlined backends are critical to ensure favorable 5G economics. One such initiative is the use of cloud-native technologies for both 5G and edge deployments, but AvidThink’s analysis from conversations at Mobile World Congress (MWC) and Linux Foundation’s Open Source Leadership Summit (OSLS) indicate we still have a long journey ahead of us.
When network functions virtualization (NFV) was first proposed by leading telcos back in October 2012 and subsequently housed under the ETSI NFV ISG, the focus was around virtual machines (VMs). In particular, OpenStack was the open-source darling of the NFV movement. OpenStack, as the virtualized infrastructure manager (VIM), coupled with the Linux KVM hypervisor running on white box servers was the perfect model for NFV infrastructure (NFVI). All that was needed was a management and orchestration (MANO + NFV-O) system, perhaps an SDN controller, and a simple porting of physical network functions (PNFs) into virtual network functions (VNFs), and we would call it a day.
Of course, in the ensuing six-plus years, we’ve learned a lot more, and there’s much unrealized on the NFV front. NFV performance is still a pain-point and on-boarding VNFs continues to be a challenge. Multivendor interoperability has not materialized, and orchestration problems continue to plague us. Certainly, there have been successful deployments of NFV and telcos continue to tout their success in virtualization—virtual EPC (vEPC), virtual IMS (vIMS), SD-WAN and vCPE—but many of the original goals of NFV have yet to be achieved.
Nevertheless, telcos are attempting to slough off half-successful NFV implementations at both the core and edge and leapfrog their way into a new round of virtualization, in the hopes that they will succeed now where they haven’t before. As a result, we now have the concept of cloud-native network functions or CNFs. The CNF concept is championed by the Linux Foundation (LF) and particularly by its Cloud-Native Computing Foundation (CNCF), a group focused on containers and cloud technologies under the LF.
Tapping into current hype around the use of low-overhead, fast to spin-up environments that provide just enough isolation with a bunch of attendant software-development benefits, e.g. continuous integration/continuous deployment, the CNF movement intends to transform network functions into collaborating collections of microservices that combine to provide network services. These container-based microservices can be orchestrated by Kubernetes, Google’s hottest recent contribution to open source, which can manage container clusters at large scale. This services-based approach is also an element in the 5G SBA (service-based architecture) framework, critical to network slicing and other services. And my conversations with vendors and telcos at MWC and LF OSLS usually had them touting container infrastructure as a critical element of all their 5G and edge initiatives, with Kubernetes as the new open-source star, shining brightly in everyone’s marketing-architecture diagrams.
The CNCF’s latest evolution of NFV from physical to a cloud-native world is depicted in the following diagram from the CNCF:
What’s smart about this move is the use of KubeVirt and Virtlet to accommodate VM-based VNFs within a Kubernetes container-centric environment. As a result, OpenStack is now missing from the diagram on the right. Also note that VMware is missing from the middle diagram. VMware’s absence might be explained by its push via both Pivotal Container Services (PKS) and the Heptio acquisition into the world of Kubernetes. Despite the absence of OpenStack on the right, the reality is that most present-day NFV deployments I’m aware of are still based on OpenStack or VMware with limited production-deployment on containers. And what’s interesting is the inclusion of ONAP as part of past architectures, which might be a stretch given the current state of ONAP deployment. Note also that while a smart move, KubeVirt and Virtlet are works-in-progress and have limitations in handling multi-VM systems. I’m probably nit-picking here and expect the diagram will likely evolve over time as Dan Kohn and the CNCF members develop the vision and mature the technologies.
Regardless, to achieve the vision, it’s not as simple as shoving existing VMs into containers as we tried to do with shoving physical functions into VMs during the early days of NFV. There’s significant work to be done.
To gain the full benefits of a cloud-native infrastructure, network functions will need to be refactored to fit microservice architectures, and vendors will need to figure out how to deploy collaborating microservices with a Kubernetes scheduler to enable orchestration and auto-scaling of different services. It’s not clear we know how to do that today.
Beyond that, it’s also not clear how a service-level orchestrator like ONAP will collaborate with the scheduler within Kubernetes to ensure correct deployment of network services as collaborating microservices. Similarly, the telemetry from microservices to drive scaling at the services-level needs to be worked out.
And finally, as vendors break VNFs down into microservices, we’ll come to realize similarities between components—after all, how many packet classifiers do you need on your CNF cluster? There’s likelihood that this new level of disaggregation will upset the current value chain, reducing the value for some while creating opportunities for others.
In the end, we at AvidThink submit that there’s plenty we don’t know yet about containerization and CNFs for 5G and the Edge. There’s work to do as a community in figuring out the implications and ramifications of the move to CNFs. Time to roll up our sleeves!
Roy Chua is Founder and Principal at AvidThink, an independent research and advisory service formed in 2018 out of SDxCentral's research group. Prior to co-founding SDxCentral and running its research and product teams, Roy was a management consultant working with both Fortune 500 and startup technology companies on go-to-market and product consulting. As an early proponent of the software-defined infrastructure movement, Roy is a frequent speaker at technology events in the telco and cloud space and a regular contributor to major leading online publications. A graduate of UC Berkeley's electrical engineering and computer science program and MIT's Sloan School of Business, Roy has 20+ years of experience in telco and enterprise cloud computing, networking and security, including founding several Silicon Valley startups.
Industry Voices are opinion columns written by outside contributors—often industry experts or analysts—who are invited to the conversation by FierceTelecom staff. They do not represent the opinions of FierceWireless.