Mobile network virtualization is real

A couple of months ago, we had a detailed briefing from AT&T on the status of their ongoing project to virtualize their network. Given the size of AT&T's undertaking (they have a rather large network), you would be forgiven for thinking that AT&T had taken a few baby steps to date and had a decade or so to go before the network was rebuilt/redesigned using virtual principles.  As we shall see, this is not the case - AT&T is making real progress.

The concept of network virtualization for the mobile networks has been around for a few years.  While numerous trials and proof-of-concepts have been conducted by operators around the world, showing solutions for the RAN and EPC, the majority of commercial deployments to date have been for small segments of an operator's business or for MVNOs.  Of course, given the potential benefits of re-architecting the mobile network using a virtualized approach, the vast majority of mobile operators have investigated the approach and are in the process of deployment.  But few details have been made publicly available.

As we have discussed before in this newsletter, one of the biggest challenges with moving to a SDN/NFV architecture is shifting the employee mindset and skill set to that of a software company.  AT&T noted this first in their briefing; they are now thinking of themselves as a software company, not a telco.  Obviously, for those of us who have grown up with networks, class 5 switches, TDM and the like, this is a big shift.  Employee training has obviously had to change to accommodate this need.

To put some scale on the task ahead of them, consider that AT&T, in addition to its regional data centers, has approximately 4,600 Central Offices or local data centers, depending on your nomenclature.  While these locations originally housed local switches (and were referred to as 'switching centers' by some), their role has obviously shifted in an era of soft switches, remote radio heads and locally cached content.  Updating 4,600 of anything takes time and careful planning, especially when the revenue-generating network is still in full use!

AT&T is migrating its mobile packet core to a new architecture using standard off-the-shelf hardware; this is the main effort for 2015 and obviously impacts a significant amount of AT&T's mobile traffic.  To date, in line with the approach taken by other mobile operators, AT&T has used virtualized packet core solutions to support its MVNO and connected car businesses.  This will obviously change in 2015 as more of AT&T's core revenue-generating traffic shifts to a virtualized EPC.

In 2016, the focus of the virtualization project will shift to the backbone.  The goal is that, by 2020, approximately 75 percent of AT&T's 150 network functions will be cloud-based and under NFV control.  At the end of 2014, approximately five percent of the network functions were cloud-based.

AT&T expects much of the benefit from the SDN/NFV architecture to come from reduced operating expenses. Obviously, this is critical to remaining competitive moving forward since mobile revenues are not likely to (some would say 'certain not to') increase at the same pace as the demand for mobile bandwidth.  The ability to get more out of the network at lower operating cost (and therefore at a lower effective 'cost per GB delivered') is therefore paramount to AT&T being able to maintain - and grow - margins.  AT&T's reduction in capital spending has been well publicized (and blamed for lower-than-expected results at the major infrastructure vendors in the final quarter of 2014) and it is interesting to note that in the briefing the company said the shift to SDN was part of the reason for the drop.

So what does this all really mean?  Simply that mobile networks are changing, and fast.  The amount of dollars spent, and where they are directed, is shifting.  Operating expenses are falling, allowing the large operators to offer a lot more bandwidth for slightly more dollars while maintaining or growing revenues.  The ability to compete in 2019 and 2020 will not depend on the technical solutions delivered in a few years - the seeds of future competition are being sown now.