Oct 6, 2015
I moderated a workshop on the topic of software defined WANs (SD-WAN) last month and I’m continually baffled that the focus of software-defined networking remains on the hardware for so many individuals. This is clearly a hangover from the early days of SDN when many of the start ups were hammering home the point that hardware is expensive and businesses should move to a software-defined model and cut capital costs.
Unless you’re Google, Facebook, or Baidu, I believe that evolving to a software-defined model for CapEx is wrong. In fact, I’ve talked to many organizations that have told me that when they shift from a dedicated appliance to software running on off-the-shelf hardware their initial total cost of ownership goes up due to things like VMware licenses, hiring of new talent, programmer support, etc.
Now, in no way am I saying don’t do SD-WAN. In fact, I think it’s something that any enterprise with a WAN, no matter how big or small, should look at. But don’t focus on the hardware side of the cost model. In fact, if a company simply removed a dedicated appliance and replaced it with a virtual version and left everything else the same, I believe the costs would be the same or slightly higher.
What I’m saying is to be aggressive with the software-defined model but then look past what you’re actually doing with hardware versus software and reap more meaningful rewards.
One of the areas of savings that has been well-documented on this site and others is migrating away from expensive, private network circuits like MPLS, and transitioning to a broadband WAN. Depending on the size of the network and the type of connections being used, business can save up to 90% of the circuit costs. Even if the first step is to a hybrid WAN where a mix of MPLS and broadband is being used, the savings can easily hit 30% or more.
However, the big savings is on operational expenses through the use of automation tools. Looking at WAN total cost of ownership, the people costs related to running a WAN is about 55%, and could be higher with huge global networks. The big luxury that data centers have over WANs is there are always local people present. Not so with branches. Configuration changes are always done remotely and troubleshooting is nearly impossible. When I was running networks, I remember having to call local people in the branch and asking them to tell me what light pattern they saw, or trying to walk them through how to re-cable something. Remote administration of a branch isn’t easy.
Also, the deployment of any kind of new service invariably requires a site visit. Companies waste enormous amounts of time and budget flying network engineers all over the globe to deploy new routers, firewalls, or other branch-related hardware. Think about the impact of this model. Let’s assume I have only three appliances in the branch, one with a three year replacement cycle, one with a four year and one with five. Do I really want to fly an engineer out to replace appliance one in three years, then back out the next year, and then the following? Probably not. So a more likely scenario is that I’ll either choose year four and replace one device a year late and live with the consequences and then another a year early and waste money. Or I could just wait the five years and do all three and face having outdated technology. If the old stuff is security, that could be a real problem.
Here’s where SD-WAN can step up its value proposition big time. SD-WANs can bring a level of agility and automation to a network that hardware-centric appliances simply cannot. If there’s a software-defined solution in place and a virtual appliance needs upgrading, it can be done with a click of a mouse. In fact, if the right automation tools are in place, the entire network can be upgraded with very little human intervention.
If there’s a new security threat that’s emerged and a new network policy needs to be added, the entire process can be automated, saving huge amounts of time and removing the risk of human errors.
One other benefit of the software-defined model is that lower paid engineers can do many of the day-to-day tasks instead of always having to use a six-figure salaried, high-level engineer. Let that expensive resource focus on more important things and turn the mundane over to automation tools or lower-level administrators.
If evolving to an SD-WAN is now on your radar, by all means take the leap. However, make sure the focus is on service costs and, more importantly, people costs instead of trying to save a few bucks on hardware.