Tuesday, March 18, 2014

Blueprint: NFV, SDN and Carrier Ethernet Make Progress

By Prayson Pate, Chief Technology Officer and SVP of R&D at Overture

Network Functions Virtualization (NFV) and Software Defined Networking (SDN) have been getting a lot of airplay recently, and everyone is coming out with their SDN or NFV stories.  With all the airplay on Network Functions Virtualization (NFV) and Software Defined Networking (SDN), some critically important information is getting lost in the hype. NFV and SDN are not standalone technologies, and they are not sufficient for building commercial, revenue-generating services. What about the metro access network that is used to reach customers? It is Carrier Ethernet that has become the foundation on top of which all next generation services are being built, and which must team with SDN and NFV.  After all, you can’t virtualize pipes.
Each of these three technologies (NFV, SDN and Carrier Ethernet) is powerful in its own right.  What are the keys to their individual success, and how can they be combined?

Ethernet – Heart of the Access Network and the Data Center

Carrier Ethernet got its start as a cost-effective technology based on the native format for enterprise networks.  The Metro Ethernet Forum (MEF) created a standard set of definitions and interfaces to facilitate the use of Ethernet, both as a service and as infrastructure.  These definitions were initially covered by tests specified in the MEF 9 and MEF 14 standards.  Later, this standardization was expanded into CE 2.0, with increased capabilities for management, multiple service classes and features that supported interconnect between service providers.

Ethernet is also the core of cloud/datacenter connectivity.  Modern data centers are built upon a framework of Ethernet switches, both physical and virtual, and almost all data center elements have Ethernet interfaces. Going forward, Ethernet promises continuing growth, both in speed and capabilities.

Carrier Ethernet has become the foundation for all future networking services in the cloud, whether at the edge of the network or in the data center.  While necessary, Carrier Ethernet is not sufficient for constructing modern services.  The evolution of services at the edge of the network will require harmony between Carrier Ethernet, SDN and NFV.

NFV – Bringing the Cloud to the Network

NFV is a concept introduced by a group of leading service providers.  The idea is to achieve the benefits of the modern datacenter and to bring them to the communications network: virtualization, low cost standardized hardware, ecosystems of software components coupled by open APIs, and rapid development and deployment of services, as shown in this figure.

Service providers have embraced NFV because it promises lower costs and higher revenues through service innovation.  In particular, NFV offers the following:

  • Service providers will be able to define and deploy new services without installing or replacing network elements.  Upgrades will take the form of installing new VMs, which is a much lower impact process than replacing physical appliances.
  • Service providers will be able to define new services independently from equipment providers.  They will be able to construct a new service from virtualized network functions provided by independent software vendors ISVs) or even create them internally.
  • Finally, service providers will be able to unite the connect, compute and storage components.  Doing so will expand the scope of the types of services that can be offered today.  For example, with NFV, it is possible to couple storage utilities such as Content Delivery Networks (CDN) with private networking and cloud in order to enable applications like online education.

SDN – Control and Automation

In contrast to the service provider push for NFV, SDN came out of the universities and found its way into the modern data center.  SDN provides improvements in how cloud components can be configured, connected and operated.  SDN brings network operators a model for modernizing how they control their networks, including a movement from closed proprietary systems.  As with NFV, open ecosystems of providers and components are an essential part of SDN.

Much of the early promotion of SDN has been on the use of the OpenFlow protocol used to program the forwarding behavior of the network in a standard and open way.  By deploying OpenFlow-enabled devices, a service provider could develop new services, and even new protocols, without upgrading the network.  This OpenFlow-centric view of SDN has been hampered by its limited support of equipment in the network.  Even so, it does hold great promise for the future. In addition, alternatives such as NETCONF and RESTFUL interfaces may also be paths through which to gain the benefits of SDN.

Combining NFV, SDN and Ethernet

NFV, SDN and Ethernet are all powerful components in their own right.  This figure shows how they are being combined.

  • Ethernet and SDN currently co-exist in the data center, especially inside virtual switches (vSwitches).  Today’s vSwitches are controllable by OpenFlow to provide interconnection of virtual machines.
  • SDN and NFV currently intersect in the Central Office (CO), which is the next generation data center.  As such, it is a ripe target for early deployments of NFV with SDN being used for control.
  • NFV and Carrier Ethernet can be combined to provide some interesting applications, such as Virtual Service Edge (VSE).  VSE is the replacement of dedicated customer-located appliances (e.g. routers and firewalls) with software-based applications running on a standard compute platform and connected via Carrier Ethernet to the wide area network.

What Changes will be Required in Carrier Ethernet?

Combining all three technologies will get us to the sweet spot for future innovation.  However, doing so will require continued evolution of Carrier Ethernet, which has been constantly growing and changing to meet the evolving needs of its users. What are the next changes that are needed?

Better Coupling with SDN

As described above, vSwitches currently provide a virtual way to connect cloud components under the control of OpenFlow.  However, OpenFlow provides a degree of arbitrary flow identification that is often not needed, and which is not compatible with much of the deployed base of equipment.  For example, OpenFlow can define a flow based on parameters such as MAC addresses and QoS markings.  A simpler forwarding model and standard focused on Carrier Ethernet virtual connections would facilitate the use of SDN principles with existing infrastructure. This leads to the next area for growth.

Simpler Standardized Interfaces

Because OpenFlow can be used to control arbitrary flows, its control protocol must be equally flexible.  What is needed is a simpler model for an API and data structure that matches current capabilities.  The MEF has an opportunity to drive innovation and consistency related to controlling Carrier Ethernet infrastructure.  In particular, the definition of a schema to match the current CE 2.0 capabilities would be very useful.

Definition of New Services to Support Cloud

Despite its power, Carrier Ethernet has some limitations, especially with respect to VLAN and MAC scaling.  These limits have driven work on new protocols for network virtualization such as VxLAN and NVGRE.  These technologies allow the construction of overlay networks to decouple the VLANs needed by virtual machines from those used to provide the underlying connections.  There is still disagreement, however, about exactly what new capabilities are needed, causing divergence in how network virtualization is being implemented.

Carrier Class Orchestration

In the data center, orchestration refers to automated provisioning, coordination and management of physical and virtual resources.  In the data center, resources are homogeneous and fungible, and we can assume fat pipes for connection.  This is not the case for application of SDN and NFV to the access network using Carrier Ethernet.  We need orchestration that can take into account such parameters as cost and availability of resources, location and latency, and available bandwidth.  In addition, elasticity and multi-tenancy are key attributes of a carrier-class orchestrator.


By combining the powerful technologies of NFV, SDN and Carrier Ethernet, service providers will be able to reach their goals of modernizing their networks, cutting costs and, most importantly, accelerating the creation and deployment of lucrative new services.  Continued innovation is required for Carrier Ethernet to be able to support this evolution.  Adding new capabilities will enable efficient expansion of new services beyond the data center into the Carrier Ethernet-powered metro service edge.

About the Author

Prayson Pate is Chief Technology Officer and SVP of R&D at Overture, where he is also a co-founder. Prayson is a technology leader and evangelist with a proven track record leading teams and delivering products. Since 1983 he has been building Carrier Ethernet and telecom products for service providers and network operators around the world - both as an individual developer and as a leader of development teams. Prayson spends much of his time driving adoption of Overture's new Ensemble Open Service Architecture, which includes aspects of automation, virtualization, SDN and NFV. He has a BSEE from Duke, an MSECE from NC State and is the holder of nine US patents.

About Overture

Overture is the preferred provider of Carrier Ethernet solutions for the metro service edge. By leveraging Overture’s Carrier Ethernet expertise and its new Ensemble Open Service
Architecture™ for software-defined services, network operators and service providers worldwide are maximizing operational efficiencies and introducing new revenue-generating services on a scale never before possible. Overture is headquartered in Research Triangle Park, NC, with sales offices around the globe. For more information, visit http://www.overturenetworks.com