Let’s Clear Up Some Misconceptions – Is SDN Relevant to the Enterprise WAN?
June is summer weather in the San Francisco Bay Area, but quite different from the June I was used to in Boston. A common misconception around Mark Twain and his relationship with San Francisco summer is that he never said “The coldest winter I ever spent was a summer in San Francisco.” But he did say: “Cold! If the thermometer had been an inch longer we’d all have frozen to death.”
June is always exciting here and in Europe. NBA playoffs are on and the Giro d’Italia just ended with the Italian Vincenzo Nibali winning it emphatically. The San Antonio Spurs have been so good but King James brought the Heat back from a certain death in Game 6. Game 7 decides it all!
Equally exciting is the buzz in networking circles, especially in the Bay Area, around Software Defined Networking (SDN) and how it is potentially commoditizing networking infrastructure. However, just as we cleared up that misconception about Mark Twain, I’d like to clear up some points around WAN challenges with cloud migration and how SDN might be applied to overcome these challenges.
In this blog post I will discuss the challenges in the Enterprise WAN and relevancy of SDN in overcoming these challenges. In part 2, I’ll cover how the Cisco ONE Enterprise Networks Architecture addresses these WAN challenges.
Enterprise WAN Challenges
As enterprises are consolidating their IT infrastructure in private cloud (enterprise data-centers) or public/hybrid clouds they’re realizing massive economies of scale in application deployments. Further, they’re taking advantage of XaaS (Software/Infrastructure as a Service) offerings from Cloud Service Providers with Pay As You Go models that increase the speed of deployment and the agility of their business critical applications. This is a major shift in how applications are now being delivered over the WAN to their end-users in branch offices and on mobile/BYOD devices. IT consolidation and virtualization in the data-center are placing a lot of requirements on the enterprise WAN. Business agility, end-user and customer application experience are imposing critical requirements on WAN. The major challenges that enterprises are facing with cloud migration are:
- Application Visibility and Control: lack of visibility into applications running in the branches and being served over the WAN from a consolidated data-center.
- User experience: End-users accessing applications and enterprise customers accessing applications on BYOD devices experience high latency due to congestion on the WAN. This negatively affects the employee productivity and customer loyalty.
- Security: Enterprises are having to hairpin all branch traffic through the data-center, where security policies can be applied. This increases congestion on the WAN link with non-business critical applications potentially consuming a significant portion of the WAN bandwidth.
- BYOD traffic visibility: With employees and customers increasingly accessing applications on BYOD devices over the wireless infrastructure, enterprises have no visibility into these applications running and cannot adequately prioritize their WAN bandwidth consumption.
Is SDN a solution to these WAN challenges?
SDN is defined by Open Networking Foundation (ONF), as:
In the SDN architecture, the control and data planes are decoupled, network intelligence and state are logically centralized, and the underlying network infrastructure is abstracted from the applications.
Additionally, the ONF SDN architecture is defined as shown in Figure 1 below.
The SDN Control Layer (an SDN Controller) has centralized control and network state which are decoupled from the network devices. This implies that network devices are programmed, through open APIs, with the network state and services as determined by the control layer. Such a tightly coupled loop between the control layer and the network devices is not applicable to all segments of the network.
SDN as defined above applies to the physical and virtual networking infrastructure in data-centers where the network devices are mainly L2 switching devices. These devices are both physical devices, such as the Nexus switches and virtual devices such as the vSwitch, Nexus1000V switches. SDN was designed to rapidly provision networking infrastructure, both physical and virtual, to keep pace with the speed of being able to instantiate applications and services in the data-centers.
However, extending this concept of SDN to the broader Enterprise network, specifically the WAN connecting the data-centers to various branches and mobile users, is creating massive confusion. Plenty of questions arise:
- What is the SDN control layer for the WAN networks? Is there going to be real-time control of the routes and flows similar to what an SDN controller does with the data-center L2 network?
- Are branch routers going to be “dumb” network devices?
- Do we need routing/L3 at all if SDN centralizes the control layer?
In my next blog I’ll address these questions. Thanks for reading and stay tuned for the next blog in the series. As always, any and all comments are welcome.