RAD's announcement yesterday of a new white box option for its virtual customer premises equipment is an example of how vCPE is rapidly evolving -- potentially in different ways for different operators. (See RAD Adds Whitebox Option to vCPE.)
It's a topic I expect to hear a great deal about next week in Denver, at Light Reading's NFV & Carrier SDN event. Given that vCPE was an early virtualization business case, it's not surprising to see that market develop quickly, but what is interesting is the way different vendors and service providers see that evolution.
RAD Data Communications Ltd. is a good case in point. The company was an early proponent of what it called Distributed NFV, which retained some functionality at the customer premises, even as it virtualized functions in software so they could be remotely deployed and updated. RAD's ETX-2i, launched 18 months ago, is a network termination device for IP and Carrier Ethernet services with a field pluggable x86 NFV module for hosting VNFs. (See RAD Launches vCPE Platform for Hosting VNFs, RAD Pushes Distributed NFV Forward and ESDN: RAD Rolls Out Distributed NFV Strategy.)
Some competitors to RAD (I'm looking at you, Accedian ) chose to strip down what they put on-prem to the bare minimum or look to put everything into the cloud, if possible.
With yesterday's announcement, RAD is essentially doubling down on its approach, offering a server-based white box option that it says will allow service provider customers to have the best of both the physical and virtual function worlds. RAD says it is working directly with a Tier 1 operator it cannot publicly name (oh, isn't that always the way it is?) in designing this product.
The ETX-2i with Whitebox Plus supports physical network functions such as IP-VPNs, performance monitoring, business service SLAs and policy-based traffic prioritization, but those don't use up CPU resources, which are then available to enable easy spin-up of VNFs, on a per-license basis. RAD is touting that as a platform for an unpredictable future.
A customer can start with a pure-play white box, with all functions virtualized, but when stronger routing is required -- for example -- a routing PNF can be turned up that meets performance and scalability requirements but doesn't consume the CPU, says Ilan Tevet, head of RAD's service provider line of business.
The PNFs "provide better performance and better scale than the equivalent software-based virtual functions, but they don't require any CPU, because they would run on the FPGA fabric of the RAD product," he says.
For the acronym challenged, FPGA is a field-programmable gate array, or an integrated circuit that can be programmed after manufacturing.
Service providers can also choose to use PNFs out of the gate -- or the white box? -- to meet their specific service needs. It's all about flexibility and paying for what you need now, RAD says.
"The customer doesn’t need to pay for [functions] in advance," Tevet says. "He can get solutions by activating PNFs as he wants. Or he can just have the peace of mind that he can evolve to a higher performance later. And he will be able to do so by just downloading a software license key to get more from existing infrastructures."
RAD is supporting the typical set of VNFs, including security, SD-WANs and IT services.
Virtual CPE is now well into its deployment phase and it will likely be actual deployment issues that drive this next generation of products. RAD's Tevet confirms that his company's approach is being propelled by customer demands and real-world experiences.
We'll be digging deeply into those experiences next week in Denver with AT&T, Verizon, CenturyLink, Level 3 and Equinix all on hand to share insights. It's not too late to join us, you can register here.
— Carol Wilson, Editor-at-Large, Light Reading