Attending the NFV World Congress last week gave me a chance to talk to many carriers about their migration into a virtualized network. Many have made big progress, very remarkable AT&T, Telefónica and Verizon transformations.
The congress is still small in size compared to others like the Openstack Summit, but it has grown to about over 1,300 delegates in 2016 plus 60+ exhibitors and over 30 live demos.
Carriers continue to aggressively push into NFV with NTT announcing they will virtualize 75% of their network by 2020. AT&T and now NTT Docomo and many others have either declared their virtualization goals or have begun to take active steps toward achieving them.
Simplification, automation and virtualization are words that have been used by marketers for years.In my opinion we are at the point were these concepts are not just talk and the “rubber meets the road” point is mature today. Technology decisions must be made and deployments have to happen which utilize software innovations from vendors and Open Source communities, in harmony, and only then, will provider networks transform to support the next wave of services.
Carriers are looking for "the low hanging fruits" in virtualization and many have found it in virtual Customer Premise Equipment (vCPE). CPE is the gear that carriers install in the enterprise to deliver connectivity service and acts as the traditional demarcation point between the carrier network and the enterprise network. However, CPE also has two additional characteristics that make it a prime candidate for virtualization:
It is a significant expense for carriers. Large CAPEX and OPEX is spent annually on the hardware needed for every enterprise branch location as well as the truck rolls required to install the hardware.
It is largely a self-contained function without the need to integrate into the legacy OSS/BSS systems for the carriers.
This leaves vCPE as the logical choice to save CAPEX and OPEX as a completely SW-based solution in a simple, standalone device that can be easily deployed. Multiple carriers discussed their plans to roll-out vCPE as the first NFV use case, leveraging virtual functions like routing, firewall, DDOS security, IP VPN and others. This is great news that carriers are moving forward into this great virtualization challenge and have identified their first easy win.
The other "easy win" is SD-WAN. Why SD-WAN and Why SD-WAN now?
Not long ago, IT teams had more straightforward objectives with their WANs. They needed to provide reliable connectivity between headquarters and branch offices over a combination of less-expensive broadband paths and faster, costlier MPLS links. IT would set up the networking hardware in corporate data centers and branch offices, manually configure it all, and then not touch anything unless absolutely necessary. Making any changes generally took weeks or even months to plan and execute.
Today, there are more core business applications residing in the cloud, like Office365, Salesforce.com, Google Apps, Skype, and YouTube, as well as more cloud-based security. That means it’s becoming less important to backhaul a lot of traffic to the corporate data center. There’s also the ever-increasing appetite for bandwidth, being driven by more video consumption and VoIP usage than ever. As organizations grow and change, they need to be able to provision and de-provision services with speed and agility. Meeting these objectives through MPLS based WAN is no longer viable, it is too costly and not flexible enough.
Customers want to automatically configure WAN connections across a combination of MPLS, Internet and cellular data links. They want to be able to rapidly and automatically change configurations using policies they set, and that weigh traffic priorities against the availability, cost, and performance of primary and secondary links. They also want to be able to do all this centrally using software, not hardware. In other words, they want SD-WANs. There are multiple approach we can take to SD-WANs but I will cover these in a separate article.
Besides cost savings, automation, SD-WAN and vCPE, the other star topic was resiliency and reliability.
In most sessions at NFV World Congress, there was something to be said about Availability, Reliability or Resiliency. Although there is a working group in OPNFV on availability, it is still an open area of concern.
I want to point out that these concepts are related but don't have the same meaning.
Availability is the percentage of time an equipment (or a network) is in an operable state calculated as percentage Availability = Uptime / Total time.
Reliability is how long a system performs in its intended function before it fails. It is typically measured in Mean Time between Failures (MTBF) = total time in service / number of failures.
Resiliency is the ability to recover quickly from failures, to return to its original form or state just before the failure, usually measured in seconds.
Therefore, a Highly Available (HA) system may not be Highly Reliable (HRel) or Highly Resilient (HRes). For example the telco "Holly grial" of 4 nines (99.9999%) availability means the equipment (or network) is never inaccessible for more than 5 minutes in a one year period, which is pretty amazing and borderline perfect. But what happens if every week it fails for just a few seconds and in the process causes every session to be initialized? How long do these processes take to start again? You now have a highly available system but with poor reliability and resiliency – this is not good for any telco. Telcos that understand the meaning of carrier-grade appreciate stateful Fault Tolerance, which provides HA, HRel and HRes. This means when a fault happens, the applications and “system” continue to run (in a secondary entity) with the same states. This is what it means to be “carrier-grade.”
Keep tuned for more news and info on NFV and SDN
