AT&T Personifies Slow Crawl to SDN Ubiquity

The Wall Street Journal reported in a “CIO Journal” blog post that AT&T intends to virtualize 75% of its network by 2020, with “very specific operational planning,” according to SEVP of Technology and Operations, John Donovan. 

Why would AT&T push off SDN so far into the distant future? Then again, 2020 is only five years away. Five years is a relative timeframe in IT, and I think the length of AT&T’s transition not only underscores the size and scope of AT&T’s networks but also the caution to which they’re giving the task. This is understandable considering the complexity of managing SDN. 

Donovan says the motivation for AT&T is to reduce capital expenditures while increasing capacity in the network. In our recent survey of service providers, more than 40% said reducing costs is their number one driver for deploying SDN, compared to only 17% in 2013. The biggest drivers cited however – which corroborates AT&T’s desire to increase capacity – are improved agility and supporting new services such as cloud, big data applications, and mobility. 

According to the article, it took AT&T “…18 months to deliver more capacity to customers in a complex network, but AT&T’s goal is to deliver it in minutes.” In addition, a Network World interview with AT&T SVP of Architecture and Design Andre Fuetsch revealed that AT&T’s Network On Demand service allows for clients to add more bandwidth or change quality of service between sites by themselves, but that change takes days or weeks. With a SDN-based infrastructure, that change may soon take only seconds – a competitive advantage and a game changer when it comes to capacity planning. 

To that end, 53% of the service providers we surveyed said they have some production SDN deployed (compared to only 19% in 2013). However, nearly 100% said that SDN requires new management tools, and that some of their existing tools will not work with SDN. In addition, more than half are concerned about lack of industry standards and SDN complexity. 

Indeed, we’ve yet to reach the point at which SDN has reached a “plug-and-play” level of accessibility, because traditional management tools are not up to the task. In the WSJ article, Donovan said that AT&T is working with open source groups such as OpenStack and OpenDaylight to “get some new thinking to create dramatic change in its software design and for economic reasons.” 

Network World reported a few months ago that AT&T was working with NTT, Ciena, Fujitsu, Intel, NEC, et. al. through the Open Networking Lab for the development of Open Network Operating System (ONOS), which effectively does the same job as OpenDaylight. In fact, AT&T is investing $1 million per year for five years in ONOS. ONOS’s developers believe that there would be less vendor-lock with ONOS than with the OpenDaylight project. 

Cisco and IBM (traditional networking hardware vendors) founded OpenDaylight, while software-oriented companies such as Google, Facebook, Yahoo, Verizon, and Microsoft support the Open Networking Foundation, which favors ONOS. 

As we’ve said many times here, SDN management needs to catch up to its promise. AT&T has thrown down a gauntlet of five years. However they get there, we hope they are successful in using SDN to deliver a new level of service to customers. For our part, we are on track to deliver SDN management in our product this year.