Reconciling the carrier SDN paradox

How does advanced technology go mainstream? In the article below, Kevin Fogarty deals with this (new) age old question in the context of network virtualization in telecom carrier and service provider networks. If SDN for the enterprise is still a puzzle for many, complexity in these environments – due, as Fogarty points out, to sheer WAN size, and the service provider’s need to support multiple client network types and legacy systems – means the challenges for this segment are legion. Ironically, these very challenges mean that this group is the one that could most benefit from virtual update, but, he explains, it is held back by a lack of standards for carrier SDN and NFV, combined with a shortage of skills, reinforced by a reluctance to invest in required software and hardware upgrades in a market that is undergoing transition. For the specialist, this discussion has much to offer: Fogarty delves into murky waters such as “clarifying where NFV stops and SDN starts” and how these capabilities “intersect” with related industry standards and open-source efforts. But it’s also important reading for the enterprise, which has much to benefit from carrier implementation of virtualization platforms, not the least of which is more flexible and more cost effective access to cloud resources. As with any early market technology, first adopters have absorbed this message and the potential competitive advantage overcoming fear of change can bring – details on how innovators are working to deliver new virtualization services to customers are introduced here, and will be more finely drawn in a companion article coming soon. Stay tuned… (ed.)       

Software defined networks have been at or near the top of IT-industry priority lists for the past three years, but the number of implementations remains small and the base of potential customers is split half-and-half on the question of whether to use it at all.

Despite what even they call the vast potential to create and sell new services to existing customers, telecommunications providers that sell most of the WAN connections and services in North America are even more cautious about jumping into software defined networking (SDN) than their potential customers.

From a customer’s perspective, SDN-enabled wide-area networks (WAN) would allow IT managers to define connections between corporate facilities as easily as they would define SDN links between two servers in a data centre. It would allow one network operations center to easily see and control all the company’s network links, create load-balancing rules to improve performance and to save money by routing data for updates or data replications on the cheapest available connections.

It could also provide single sign-on that would activate the same network context and user profile anywhere on corporate assets – meaning that a business-tripping executive could log in at a company office in California and automatically have access to the same servers and applications she uses in her home office in New York.

Carrier SDN works, in theory, the same way enterprise SDN does: by lying to applications and servers about the physical layout of the networks they’re connected to, then telling routers and switches what actual route network traffic should take in order to arrive at the destination to which applications and server think they were sent.

Carrier networks are exponentially more complex than enterprise networks simply because of their size. They are made more complex by high quality-of-service requirements,  the need to support legacy systems from as far back as the ’60s for some customers and such a wide range of options for type of network that, on metro networks, it’s not unusual to see service edge routers, carrier Ethernet switches and routers, SONET/DDH MSPPS and DWDM in close proximity, according to a 2012 white paper from Heavy Reading describing SDN plans for metro networks.

More than half (53 percent) of service providers have some version of a production SDN network up and running this year , according to a November survey of 60 carriers by network route-analytics company Packet Design. That’s more than twice the 19 percent that said the same last year.

But the number of those who said a lack of industry standards is a problem also doubled, from 26 percent in 2013 to 56 percent now.

Nearly 70 percent said they don’t have the skills or training in-house to build or manage an SDN network – a problem affecting both users and carriers partly due to a shortage of training from vendors, partly due to the hesitation of network operations staffs to trust networks controlled by software, according to a March 2014 analysis in Network World.

Despite their hesitation, WAN and telecom carriers do provide some SDN services, have promised since at least 2013 to expand those services and add new ones based open-source SDN technology as well as data centre networking and cloud-connection services.

To do that carriers would have to launch expansive upgrades and replace aging hardware that supports legacy WAN protocols with newer gear that also supports long-distance IP and SDN.

During the past year, however, carriers have become so leery of rapid changes in SDN products and protocols that they’ve actually reduced the amount they spend on new routers and switches, according to a Dec. 1 report from Infonetics.

“Carriers remain cautious about investing in equipment and software that might need to be replaced in the future,” Michael Howard, principal carrier analyst at Infonetics said in a release announcing the report, which referred to the trend as carriers’ “SDN hesitation.”

Telecoms don’t seem to think customers are going to buy fewer network services, apparently. Spending on Carrier Ethernet switches shot up by 32 percent during the third quarter of 2013 and 14 percent during the past 12 months, even though drops in spending on core and edge routers dropped enough to bring total spending down 3 percent for the third quarter.

In addition to the potential riches they see in SDN, carriers also need to expand their capacity to keep up with growing volumes of Internet traffic, which threaten to saturate the capacity of “aging core router platforms” on the Internet backbone and on carrier metro networks, to which they’ve relocated some core routers to keep up with growth in capacity, according to Dell’Oro Group analyst Alam Tamboli in a July 29 report.

AT&T, Japan’s NTT Communications and a few other carriers have pushed forward quickly with SDN initiatives to help manage their own networks and to provide smart cloud-connection services to customers.

Most, however, are nursing along the high-capacity routers that would carry the heaviest load and have to be replaced in the greatest numbers while they mull over what mix of SDN and Network Function Virtualization services each will need to deliver, how much upgrade would be required and how, exactly, to go about it, Howard said.

Carrier executives also seem distrustful of a movement that, aside from its claims of cost-cutting and increased efficiency, often criticizes existing carrier networks and business rules for being overly complex and expensive, and which censures telecom managers for being unwilling to modernize either their thinking or their networks.

Verizon Communications, Inc., for example, has a “strong program to leverage SDN and NFV,” but some senior-level executives are worried about the “operational problems” that may result when lower-level network functions are virtualized and it’s not clear which vendor might be responsible for a specific problem, Kyle Malady, Verizon SVP of global network operations, told Light Reading in June.

Network Function Virtualization (NFV) focuses on virtualizing lower-level services like DNS and DHCP rather than the content- and application-level functions that SDN takes in hand. It is often referred to separately by carriers, which often view the functions NFV addresses as more fundamental to the health of their networks than the full range addressed by SDN.

“We do not expect NFV to replace the core and access components of our global network,” BT opined in an essay and white paper that raised red flags about the extent to which virtualization based on generic x86 PC and server architecture could replace, or even equal the performance and quality of service telecom companies get from the specialized hardware and custom-spun ASICs carriers use for their highest-performance networking hardware.

Carriers aren’t sold on the idea that SDN will save them any money on hardware, because they have to vastly overprovision their networks in order to get the reliability required of a phone service but not of most things based on x86 servers, according to carrier technical experts speaking on a panel at last year’s Digital Disruption conference in October, 2013 in San Jose, as reported by Light Reading.

A Telecommunications Standards Institute (ETSI) announcement in March did break down barriers and get the carriers’ technical staffs talking, but they haven’t been talking long enough to make the much progress, according to Sprint senior network architect Fred Feisullin.

ETSI published its first set of NFV specifications in October of 2013, and announced the completion of Phase 1 and the start of Phase 2 of the specification group’s mission on Dec. 4. The announcement praised the group for tremendous progress, but listed some questions as needing to be answered during Phase 2 that seem too fundamental to have been open until now – clarifying where NFV stops and SDN starts, for example, and how either one “intersects” with related industry standards and open-source efforts.

Not all carriers are so virtually cautious. AT&T announced in March a radical restructuring plan aimed at doing “for the wide area what the world has done with the data centre,” John Donovan, AT&T’s senior executive vice president of technology and networking operations said in a keynote at the same Open Networking Summit conference at which Google unveiled its Andromeda SDN infrastructure.

In pursuit of a future network it calls the User-Defined Network Cloud (Video), Donovan promised to build a new network rather than overlay SDN-enabled hardware on top of the old. “We’re rewriting the book and setting a new foundation on how businesses manage their network services,” AT&T senior vice president of marketing and global strategy  Roman Pacewicz wrote in a Sept. 16 blog announcing a pilot test of the SDN network at the University of Texas at Austin. The service is due for commercial availability before the end of the year.

The end result will be an SDN network built on top of a series of NFV-based technologies called the  “User-Defined Network Cloud” that will be based on open-source standards, will allow users to provision their own services, create their own networks or start service by installing software instead of wires. It will also be scalable enough, Pacewicz wrote, to handle vast increases in data volumes that will push more than a zettabyte (1000 exabytes) through IP networks during 2016 – an increase that will include  11 times as much mobile data traffic in 2018 and video traffic that will account for 79 percent of data streamed to consumers by 2018, according to a comprehensive but somewhat breathless projection in Cisco Systems Inc.’s Visual Networking Index report for 2013.

An Alcatel-Lucent SDN partnership called Nuage Networks has also introduced a drop-in SDN service it provides to telecom carriers, which can give new customers a pile of software that will set up an instantiation of the company’s Virtualized Services Platform data centre VPN  to connect their sites with an SDN based in a Nuage data centre, according to a Nov. 2 Light Reading story. The all-software setup is partially an alternative to having a technician come out to make the connection, but the service itself is a virtualized but direct connection to Nuage’s OpenStack-based cloud, and is one of a whole  series of cloud, SDN virtual-everything-based services Nuage has been providing for data centres since the spinoff was formed in early 2013.

Japan’s NTT Communications announced in October that it had made a big leap into SDN, committing to a series of upgrades that would allow it to provide full-scale SDN services to manage web and cloud-based networks, VPNs and virtual infrastructures.

The upgrades will be free to Japanese customers.

NTT said it would also launch a service in Europe and in the US that would allow customers to create dedicated, VPN connections called Internet circuits into its cloud infrastructure. The service would allow customers to provision their own connections, change bandwidth levels or change their IP address structures at speeds ranging from a guaranteed 1Gbit/sec to a “best effort” at 10 Mbit/sec.

So far, no North American carriers have followed suit.