Readers who have followed AvidThink’s research and analysis in the past will know that we have been champions of network function virtualization (NFV) since the early days. Even as NFV struggles to remain relevant in the era of cloud and containers, we continue to believe in its original promise. At the same time, we have been critical of the execution of the NFV vision, and have laid out in previous posts the missteps that we, collectively as an industry, have taken along the way.
The premise behind a software-centric, virtualized, telco infrastructure running on commodity hardware, one capable of running multivendor network functions and coordinating across multiple domains, is still valid. And, more importantly, it is still a laudable and important business goal. Regardless of whether this infrastructure is virtual machine-centric (the reality today) or cloud-native and container based (a more appropriate goal for the future), a software-defined telco infrastructure can still achieve lower operational costs, improve agility and flexibility, provide faster recovery from failures, reduce complexity, and might also provide hardware savings.
Carrier SD-WAN: Round one
In the early days of telco NFV, there were a few emergent, hot use cases, including virtual IP Multimedia Subsystem (IMS), virtual evolved packet core (EPC) and virtual customer premises equipment (CPE.) While vIMS and vEPC were rolled out and continue to be rolled out today as part of foundation-laying for 5G deployments, the vCPE use case morphed. While NFV was in its infancy, carriers tried to figure out the right vCPE architecture: how much of a network function, and how many network functions, should run in a remote CPE versus at a point of presence (POP), central office (CO), or cloud data center? In the meantime, a renaissance at the enterprise edge elevated SD-WAN to a pre-eminent enterprise use case. Soon, tier-one carriers were falling all over each other to figure out how quickly roll out SD-WAN services for their enterprise customers.
Carrier SD-WAN: Round two
After achieving initial SD-WAN success, telcos realized that they couldn’t afford to be held hostage by successful SD-WAN vendors. With Cisco snapping up Viptela, VMware buying VeloCloud, and Nokia owning Nuage Networks, telcos needed leverage against these major networking players. SD-WAN was increasingly becoming an enterprise edge platform on which value-added services could be deployed, and telcos wanted to make sure they held control of that platform. And so, we embarked upon Gen 2 SD-WAN architectures, which involve telco-selected, telco-controlled universal CPE (uCPE) running a flavor of an operating system (vCPE OS), as well as a choice of one or more SD-WAN VNFs.
The goal behind these Gen 2 SD-WAN deployments is to swap out one SD-WAN vendor for another if needed, and therefore reduce the leverage of any single incumbent SD-WAN vendor. Whether this move will actually succeed remains to be seen. After all, the reality is that SD-WAN isn’t of sufficient maturity and consistency that one vendor can seamlessly be swapped for another without administrator retraining, policy and configuration remapping, automation reprogramming, and significant system reconfiguration. Even in a managed SD-WAN deployment with limited enterprise access or visibility into direct hands-on SD-WAN configuration, the telco’s managed services team will still have to go through significant efforts to swap in one SD-WAN vendor for another.
Regardless, telcos are building out infrastructure to manage and host multiple VNFs on uCPEs, as well as VNFs on POPs, COs, and cloud data centers, in service of the SD-WAN use case. Telcos are also offering SD-WAN coupled with value-added VNFs like next-gen firewalls (NGFW), WAN optimizers, and virtual routers. Some of these are to meet enterprise compliance needs, adding previously certified compliant firewalls and other security devices to the SD-WAN deployment, while others are meant as additional value-add services.
All roads lead to NFV
In early NFV demonstrations, we “oohed” and “aahed” over self-service marketplaces for a rich set of VNFs. Unfortunately, many of those demonstrations never achieved production roll-out or any significant market success.
Ironically, with SD-WAN, an arguably enterprise-led initiative, we’ve now come full circle. These Gen 2 SD-WAN deployments are morphing into fully fledged NFV deployments, with orchestration across multiple vendor SD-WAN environments, along with integration of other VNFs like NGFWs, vRouters, etc. Even better, these deployments have also expanded to orchestrate across underlay circuits as well as overlay SD-WANs. And we’re now seeing cross-domain orchestration, tying brownfield MPLS VPNs into modern SD-WANs or modern SD-WANs into express lanes to cloud IaaS, PaaS, and SaaS for improved performance and quality of service.
Further, as we have discussed in the past, with 5G FWA on the horizon, these SD-WAN-driven NFV infrastructures could expand to cross into orchestrating the 5G FWA domain too.
In many ways, the SD-WAN wave has become the catalyst for larger scale deployment of NFV: uCPE, POP, CO, cloud-based NFVI with VM-based and container-based VIM layers, cross-domain orchestration featuring NFV-O, VNFM, and SDN controllers. So perhaps, SD-WAN is the panacea we have all been looking for — it solves many enterprise headaches, breathes life into carrier NFV initiatives, provides brownfield MPLS with a second chance at life, and bridges to the 5G future with FWA. It slices, it dices, and if you configure it just right, it’ll make pasta too. What’s not to like?―Fierce Telecom