It seems that every day in telecom circles we see new press announcements that extol the virtues of network functions virtualization (NFV). There are good reasons for this, of course, given NFV has entered the implementation phase.
Nevertheless, it's also fair to say that behind the scenes, conversations that NFV is facing deployment obstacles continue to bubble up as well.
As a result, unless you are committed to following the twists and turns on a daily basis, it can be difficult to debunk some of the urban myths and establish a clear demarcation line between virtualization fiction and non-fiction. There are several factors for this, but perhaps the old adage, that the truth lies somewhat in the middle, is most valid here.
To be clear, in the spirit of full disclosure, I am on the pro-NFV side of the argument. I believe the cloud architecture is not only disruptively elegant and constitutes the reference architecture of the future, it also aligns with the low latency, application-everywhere expectations of end users.
But I am also a pragmatist, so let's spend some time and separate fact from fiction in a number of key areas; starting first with platform requirements.
Generation X requirements
One of the driving principles of NFV is to achieve functional separation of hardware from software so that applications can run in a totally independent environment, thereby breaking down vendor lock-in scenarios and lowering capex-driven procurement costs. And this is still the promise of NFV, but it's also fair to say that probably not all applications will ever run exclusively on COTS-based x86 servers. The truth, I believe, is that some edge functions including those that perform transcoding, encryption and control plane functions may never be well suited to an x86 platform given their unique packet processing requirements. The other factor to consider is that network operators are well versed in application-specific hardware and given they will be likely be operating in hybrid mode, leveraging a common platform that bridges both domains and uses common OSS/BSS tools has merit.
Scale simplification and border security
Another factor driving the widespread adoption of porting virtualized applications to x86 servers is the belief that this will achieve true elastic, cost-effective scaling and the benefits of NFV in general. Still, it's also becoming clear that simply porting applications to x86 for a number of these highly specialized functions will not deliver significant savings or even performance improvement levels anticipated. As a result, for certain functions, especially those that run at the edge such as session border controllers (SBCs), the evolution and future development cycles will entail more than simply porting SBC software to x86. In fact, I believe as we see the shift to more distributed edge computing models, it will be imperative to optimized functions like SBCs specific to the ever greater role they will take on to enhance real-time communication (RTC) application performance and manage security at the network border.
Policy and analytics
Both policy control and analytics have established themselves as important but independent telecom network technologies. However, over the past 18 months it's become increasingly apparent that both are now functional requirements for any successful NFV implementations even though as an industry we still classify the relationship as only loosely aligned. My rationale for adopting this stance is that I believe the cloud has expanded the concept of policy control from a Layer 4-7 to an end-to-end Layer 2-7 relationship model to support security enforcement and more complex routing policy. Similarity, analytics will be vital to provide the metadata and analytical trends that policy enforcement points and controllers require to take preemptive action to manage security threats and enhance application performance and personalization.
For more discussion about where the truth lies in successfully implementing NFV and separating fact from fiction, join me for a webinar with BT's Randy Schrock and Mykola Konrad from Sonus that will document the myths, challenges, as well as the opportunities associated with migrating RTC and unified communications to the cloud. Click here to register for Real-Time Communications in the Cloud – Avoid Cloud Myths with Proven Solutions on Wednesday, November 2, 12 p.m. New York/4 p.m. London time.
— Jim Hodges, Senior Analyst, Heavy Reading
This blog is sponsored by Sonus.