Today’s WANs have become remarkably complex. What used to be relatively straightforward point-to-point links have multiplied, meshed, and hybridized, as well as gaining layers of management and optimization to overcome the inherent weaknesses of telco connections and wring more capacity from them.
At the same time, more and more users and services are moving off the private WAN and into the cloud. The enterprise has little or no control here, because while it might have SLAs for its cloud services, its ability to manage traffic is pretty limited beyond the perimeter of its private WAN.
The emerging solution for this complexity and inflexibility is, according to the Open Networking User Group and others, to adopt software-defined networking (SDN) techniques and build an SD-WAN. By moving the control logic and policy management into software, and removing (or more accurately, abstracting) the variability of the underlying hardware boxes and connections, you can layer on automation and orchestration.
Done right, the result is a WAN that is dynamic, fluid, application-aware, and able to intelligently adjust to network conditions or events. An intelligent WAN can actively measure performance on the public and private routes available to it, for instance, reconfiguring and tunneling between endpoints as appropriate.
But how and where can you implement SD-WAN? The router is a obvious place, and indeed some SD-WAN technologies do exactly this, for example sitting on top of Cisco’s policy modules to orchestrate the WAN. Other vendors offer their own ranges of SDN-enabled gateways and access points.
Indeed, at first sight SD-WAN looks to be a natural evolution for SDN vendors in general. However, WAN users will know just how different the physical properties of WANs are from LANs, and of course how much the different WAN technologies can vary within that. Even their availability can vary massively! So if you want intelligent path control over technologies ranging from T1 and MPLS to DSL and LTE, software designed for LANs may not be up to the job – it may have the intelligence and the policy management, but not the link control and acceleration, say.
SDN by itself also does little to address the other pressing need for pretty much any WAN user, which is the need to optimize the various available connections for bandwidth and latency. And at the same time, the job of SD-WAN orchestration is a natural extension of the role of WAN optimization.
The other aspect is that the modern enterprise wants – no, needs – to use both private connections and the Internet, the latter both as a backup and to reach roaming users and cloud services. After all, how much use is a super-fast company backbone if the primary applications for many of your people are now Office365 and Salesforce? A way must be found therefore to integrate and manage direct Internet access over low-cost broadband, alongside the WAN for internal applications, remote office backup, data synchronization and so on.
All of this makes SD-WAN and WAN optimization a natural fit – a dynamic duo, if you like. Either on its own is still worth having, but together they are rather greater than the sum of their parts. That is especially true if you already have WAN optimization in place, because layering SD on top – as Silver Peak does with its Unity offering, for example – adds the embedded intelligence that WANs currently lack. Even better, if the WAN optimization is already software-defined in the form of a virtual appliance, there should be no need for proprietary hardware or upgrades.
Merging the two also lets you take advantage of features that are really only practical from within the WAN. Examples might include Unity’s ability to not send cloud-directed traffic straight onto the Internet, instead routing it over the private WAN until it is as close as possible to the cloud data center where that service is actually hosted – assuming of course that its monitoring shows this to be the fastest and least congested route.
In short, WAN optimization alone will give better utilization of the enterprise infrastructure, but misses all those new application areas. SD-WAN alone can give you flexibility and automation, but will not make best use of the WAN infrastructure. Only if you integrate the two are you likely to see the best results, in the shape of an intelligent WAN that allows applications to use multiple physical paths between locations, ensuring better optimization and the optimum connectivity for applications.
This post is part of an ongoing series examining the issues facing enterprises seeking to implement a Software-Defined WAN (SD-WAN) solution, as addressed in the Open Networking User Group white paper, “ONUG Software-Defined WAN Use Case”.
Image credit: Mark Anderson (flickr) / CC-BY