SDN/NFV: More than Hot Air in the Windy City

As reported in my last blog post – SDN: Déjà vu all over again? – Packet Design was a gold sponsor of Light Reading’s inaugural Big Telecom Event in Chicago this week. The organizers claim there were about 1,500 registered attendees with two thirds of them representing network operators. There were several really interesting presentations and panel discussions, and I’ve captured four key takeaways here.

SDN is getting the buzz, but NFV is where the early action is. This theme ran through the conference and there was much discussion on the relative merits of SDN vs. NFV and whether they should be implemented simultaneously or separately. While there are clear near-term benefits – including lower capex and faster time to revenue – for operators who virtualize network functions that today run on hardware appliances, most agreed that this is a first step only. To achieve maximum efficiency and flexibility as well as deliver the best customer experience (see below), an SDN-enabled network function virtualization infrastructure (NFVI) is needed in which pools of resources can be programmatically and dynamically orchestrated to provision and de-provision virtual network functions at any time and any place. The industry is moving rapidly to implement step one and most operators say they have production deployments already, or will have later this year. While there are certainly some early production SDN deployments, true NFVIs are not yet reality.   

Operators are moving to customer-controlled service delivery. Michael Rouleau from TW Telecom highlighted the major shift that operators are making in the way they deliver services, moving from traditional service ordering and provisioning processes to an on-demand model. Consumers click to connect, use what they need and pay as they go. This has ramifications not only for the network infrastructure but also on OSS/BSS systems. More on this later. In a similar vein, Axel Clauberg from Deutsche Telekom gave a really interesting talk on his company’s transition to becoming a ‘Software Defined Operator.’ He described the simplification initiative they have to reduce the number of network technologies and protocols, which will ultimately free them to build a more flexible service delivery infrastructure. His high-level architectural view contained just three elements: an IPv6 network (with no MPLS), an NFVI cloud, and SDN. 

Can everyone just get along? Another recurring theme at the conference was the call for greater collaboration to accelerate adoption of SDN/NFV. Virtually every speaker, representing carriers and service providers as well as a number of vendors, acknowledged that new SDN/NFV technology is only part of the equation; industry players must also put aside proprietary, legacy thinking and business models to deliver a better consumer experience. David Flessas, the head of network operations for Time Warner Cable, said what is missing – the Holy Grail – is a service management view across all the different technologies involved in delivering consumer services (How long have we been saying this?). 

The good news is that industry associations provide forums for stakeholders to hash out not just the technology interfaces, but also to remove the self-imposed (competitive) business barriers and legacy cultures that prevent the creation of global interconnectivity needed to fully embrace new applications and service models. This includes the Alliance for Telecommunications Industry Solutions (ATIS), a 30-year-old North American association whose members are the who’s who in the ICT industry. 

Marcus Weldon, CTO of Bell Labs at Alcatel-Lucent, used an air travel metaphor to describe the way this global interconnectivity could and should work: We can reserve and pay for multi-hop international air travel, involving several carriers and routes (networks), classes of service, aircraft (devices) and destinations (applications), and receive the services in a relevantly seamless experience. We don’t book a flight from point A to B and then book a flight from B to C. We book and pay for A to C as one transaction. 

Existing management systems must adapt. Mr. Clauberg described his view of the SDN-based, real-time OSS that will be required at Deutsche Telekom. The OSS will maintain service models and network models accessible by the management applications and will leverage open standards, such as Netconf/Yang and OpenFlow, to manage the underlying multi-vendor infrastructure. 

This need for next generation management systems was amplified by Bikash Koley who, in describing Google’s SDN three-plane architecture (management, control and data planes), stated that SNMP-based systems can’t cut it; real-time telemetry is needed (Something we’ve been harking on in this blog for a while). Mr. Koley talked about Google’s efforts to define a high-level declarative language in the management plane through which operators can describe, in non-technical terms, the network topology and configuration requirements needed to support application demands. This ‘intent’ would then be implemented via northbound APIs (TBD) in the control plane. 

Packet Design’s CTO, Cengiz Alaettinoglu, participated in a panel discussion on SDN/NFV management and orchestration. It was one of the more lively panels, and I will cover it in a later blog.