Posted by Mae Kowalke on Wednesday, April 06, 2016 with No comments
Despite citing “software not yet carrier-grade” as a barrier to implementing SDN/NFV (in a recent ISH/Infonetics survey), operators are still moving ahead with adopting this technology—and influencing its development. That observation on the part of Michael Howard, Senior Research Director, Carrier Networks, IHS, kicked off a webinar (“Carrier NFV and SDN Lessons from Virtual CPE Deployments”) this Tuesday, setting the tone for an informative and lively discussion about how telecom operators are deploying virtualization technology.
Two main drivers are motivating carriers to adopt SDN and NFV, Howard said: service agility for quicker time to revenue, and global view of services and multi-domain, multi-vendors networks for automation. But, the transition isn’t an easy one. For example, operators are challenged with how to inter-operate physical and virtual in existing networks. Other challenges include choosing the right virtual network functions (VNFs), and navigating virtual network function (VNF) licensing.
In part because some of the deployment options operators are exploring address that interoperability question, virtualized customer premises equipment (vCPE) is arguably the leading use case for NFV at this stage in the virtualization evolution.
“Virtualization is all about flexibility—being able to consume software in a way that best meets your service requirements,” said panelist Chris Thompson, Director of Product Management, ADTRAN.
Panelist Scott Sumner, VP Solutions Marketing at Accedian, noted that since operators have spent the past couple years testing vCPE proofs-of-concept and are now making decisions about the deployment option they’ll use for commercial roll-outs, it’s a good time to step back and revisit the original reasons for adopting virtualization.
The basic tenets of traditional service delivery remain in place, Sunner noted, with new considerations layered on top. How well does the operator’s vCPE strategy address these criteria? Operators should be asking, Why are we doing this? In many cases, the answer is turning out to be less about cost savings and more about ‘softer’ benefits like service agility.
So, how should vCPE be deployed? The answer depends on the functions involved, how existing infrastructure is set up, and the operator’s virtualization roadmap.
Three options predominate: localized (aka uCPE; functions hosted at customer premises), centralized (functions hosted in provider cloud/data center), or hybrid (aka distributed; mix of localized and centralized functions hosting).
Accedian’s stance emerged during the webinar: in the long run, the centralized option is the most sustainable and provides the most benefits. But, that may not be feasible right away, and it may be a long time (or never) before some functions with especially stringent latency and security requirements can be delivered in this way. Because the market is not standing still, operators should have a roadmap regarding how they’ll host functions as technology evolves.
Panelist Richard Arthur, NFV Solution Manager, Hewlett Packard Enterprise, noted that building out a data center needed for centralized vCPE deployment is an investment, but one that will have a solid ROI if it’s well-planned.
Regardless of the operator’s strategy, any vCPE deployment should acknowledge the distinction between network demarcation and function/service demarcation, Sumner noted. The goal with vCPE is to virtualize functions; some hardware will still need to remain in place for network demarcation, although it can be reduced to a NID or even a smart SFP module.
vCPE is more complex than just software running on an x86 or similar hardware, Sumner stressed. It involves a multi-layered NFV infrastructure ‘stack,’ including hypervisor, virtualized infrastructure manager (VIM), NFV orchestrator (NFVO), and catalog of VNFs. In the long term, a well-thought-out vCPE design should make services easier to assure, manage, and deliver, but there’s a lot of complexity getting to that ‘simple.’
Howard ended the webinar with four main conclusions:
- vCPE is the #1 NFV use case
- It’s easy to test vCPE in PoCs, but more complicated for commercial deployments
- Stringent SLAs still require hardware-assisted monitoring
- Operators want vendors focused on ‘open’ (source, standards, innovation, etc.)
Watch the full webinar.
- White Paper: Decision Factors in Service Providers’ vCPE Migration
- Webinar Replay: vCPE Deployment Options
- Video: vCPE for Carrier Ethernet