Summary Bullets:
- At first sight, NFV looks to remove some of traditional benchmarks of telecom product comparison: performance and capacity metrics.
- In reality, nothing has changed. Now, it is a VNF’s capability to manage the NFVI through the VNFM that is going to count.
While not a quant house, Current Analysis is well known for both qualitative and quantitative product comparisons, and for the latter, it uses an extensive repertoire of product metrics, often somewhat disparagingly referred to as the “speeds and feeds.” While NFV is some way off, or will never happen, for some market segments (e.g., the IP and optical core; see “Specialized Network Devices Will Not Be Going Away (Anytime Soon)”), a new picture is emerging in the more NFV-intensive zones of the service and control layers.
Historically, some companies have been very reluctant to quote performance or capacity numbers, wary of unfair comparisons being made (“depends on the call mix,” etc.), while others see it as an opportunity to trumpet the power of their offering. Interestingly, vendors of both persuasions are now coming back to Current Analysis with responses such as “BHCA is now based on the performance of the COTS IT platform” or, for the same reason, “no reasonable upper limit exists for the number of simultaneous calls supported.”
So, does “unlimited” processing and storage spell the end of telecom product comparisons based on performance metrics?
As ever, the answer is yes and no. While on the one hand, NFV is refocusing attention on pure telecom application (VNF) functionality rather than the performance of the underlying platform, the challenges of maintaining high performance under load (or graceful degradation, if that’s not possible) for telecom applications is never going to go away. It’s this very issue, the devolving of that performance responsibility to the NFVI, which is now causing so much intense debate. For that reason, it’s certainly premature, if not naïve, to say that application performance and capacity can simply be substituted by the “unlimited” compute and storage resources of the NFVI.
No, NFV is not going to remove responsibility for performance from telecom application design. Rather, in NFV management and orchestration (MANO) terms, this responsibility looks to be relocated to a new entity called the VNF manager (VNFM). Instead of designers optimizing their own dedicated middleware and hardware, the hard part will now be in optimizing and orchestrating all VNF components (VNFCs) using the VNFM and the NFVI virtual infrastructure manager (VIM) to get performance. It’s not going to get any easier; in fact, it may get a whole load harder, especially as each application will be vying for a share of NFVI resources or requesting extra ones on a real-time basis.
So, what about new VNF performance and capacity comparison parameters? That’s a hard question too, and one we are actively working on in Current Analysis. However, vendors, take note: if you start to use the “unlimited” answer to the performance and capacity question on the basis of common NFV infrastructure, then be ready for the follow-up: “Ok, tell me about your VNF architecture, your VNFCs and your VNFM…”