Ramesh Prabagaran | No Jitter |
Software-defined networking has a role to play in the local and wide area — just make sure you know the distinctions.
A growing misconception is that software-defined networking (SDN) is a one-size-fits-all technology that solves a multitude of problems anywhere in an enterprise network. The LAN and WAN, however, are fundamentally different and have their own characteristics. By design, they require unique approaches to SDN.
Let’s consider what makes them different.
- Predictive versus unpredictive topology – The LAN has a predictive topology with layers of switches and interconnects, each of which aggregates bandwidth for the layer below. The WAN, meanwhile, does not. Even though sites connect with each other, WAN infrastructures use hub-and-spoke and mesh topologies, regional aggregation, traffic-engineered paths, and the like. In turn, these overlays are built on top of multiple carrier networks, each with its own topology. This leads to unpredictability in the routing of traffic from one site to another.
- Network security – In the LAN, authentication serves as the primary security mechanism; encryption is seen as unconventional. The WAN is a completely different animal. It requires authentication of end-point devices and encryption of bits on the wire.
- Unpredictable loss, latency, and jitter – These conditions are close cousins of the WAN’s unpredictable topologies. Avoiding these conditions requires use of acceleration and optimization of applications or dynamic path management based on network SLA characteristics. None of these problems apply to the LAN since bandwidth and latency are assumed to be nonissues.
- Aggregating flow information into larger manageable chunks – The number of LAN traffic flows is proportional to the number of users and applications. As a result, SDN use in the LAN is oriented on user and application populations. In the WAN, accounting for users and applications is a consideration, but operating the network based on these factors alone is impractical.
- Quality of service – The LAN implements QoS to mark traffic in a way that leads to prioritization in the WAN. Therefore, the SD LAN focuses on classification and marking. In contrast, applying SDN in the WAN focuses on shaping, prioritizing, and selective dropping. These are different functions, but usually classified together under the QoS umbrella.
- Efficient exits – One of the fundamental problems that the SD-WAN addresses is cloud-readiness and improved exit to the Internet. For example, one common enterprise concern is handling guest WiFi traffic differently than cloud-based core business applications like HR and CRM. Providing appropriate QoS for these separate classes of traffic is a priority in the WAN, while LAN architectures generally do not concern themselves with Internet and cloud exit points.
Despite all these differences, we can find some common ground when applying SDN principles to the WAN and the LAN, especially around segmentation and simplified operations. With respect to network segmentation, SD-LAN has traditionally been at the forefront, with virtual LANs, virtual extensible LANs, and overlay tunnels. SD-WAN is now catching up in this area.
So while the goal of deploying SDN, namely agility and operational simplicity, is equally applicable to the WAN and the LAN — the challenges they address in these two segments of the network are markedly different. Before you commit to either or both, make sure you understand the unique challenges of your network environment!