SD-WAN is Important for an IoT and AI Future

IoT, SD-WAN, and AI: three IT buzzwords from the recent past, the present, and the near future which, it turns out, are all inextricably linked. SD-WAN technology is going to be key to the continued success and expansion of IoT projects, while managing SD-WANs is going to be one of the key use cases for AI, in the short to medium term.

To understand why, let’s take a closer look at SD-WAN technology and some of its pros and cons.

Multiple Connection Types

SD-WAN decouples the control plane from the data plane in a network just like in software-defined networking (SDN). In a wide area context, this enables the network to be controlled and managed by software from a central location — and “the network” can include multiple connections and connection types such as MPLS, broadband, LTE or, increasingly more common, 5G.

Also read: Transforming Networks: From Virtualization to Cloudification

Application Aware

Since an SD-WAN is application-aware, that also means that applications can be made to make use of the most appropriate connections (for latency, bandwidth, or even security), and traffic destined for the cloud does not first have to be routed to the corporate data center from an enterprise branch office or home worker. Instead it can head straight to the cloud, reducing latency and corporate network congestion and increasing application performance.

Benefits of SD-WAN technology are that it can:

• Increase application performance, which is particularly important for sensitive applications such as voice, video, or perhaps live financial data feeds
• Use lower-cost links for applications that are less important or less performance sensitive
• Use lower-cost off the shelf hardware for routing, rather than proprietary routers with proprietary firmware
• Make it easier to get visibility into the network to manage it, and to make changes, from a single location. Configuration or policy changes are pushed out to network hardware
• Allow new branch offices or other locations to be provisioned and brought online much more quickly and easily, sometimes without the need for a time consuming visit from a network engineer. (In theory, at least.)

Mistakes are Inevitable

But, and this is a big but, there is no doubt that SD-WANs are complex beasts. That means that deployment, configuration, and network changes can be tricky. Mistakes are inevitable, and when they happen it can be very difficult and time consuming to figure out what has gone wrong.

So how does this relate to IoT and AI?

Increase in IoT data

Let’s take a look at IoT first. After a year in which the growth in spending on IoT projects slowed due to the pandemic, this growth is expected to increase sharply, from $749 billion in 2020 to $1100 billion in 2023, according to a Statista forecast.

That means that there are going to be a large number of enterprises introducing or expanding IoT projects, and that in turn means a huge number of new sensors generating and transmitting vast amounts of data.

Also read: The Home SD-WAN and SASE Markets are Rapidly Expanding

SD-WAN scale up

A traditional network infrastructure would likely struggle to cope with this, but SD-WAN technology can take this in its stride by scaling up relatively easily wherever it is needed, and by accommodating the performance and other needs of particular IoT applications. What’s more, SD-WAN technology can be used to protect sensitive IoT  data by imposing security controls and also through pathway isolation if required.

Where IoT sensors are remote, a popular solution is likely to involve the use of 5G networks, and the ultimate destination of the traffic may be the enterprise data center, edge data centers, or data aggregation centers hosted in the cloud.

Too complex to manage

So far so good, but now we come to the problematic part of SD-WAN, and that’s that it is, as mentioned earlier, extremely complex. So complex, in fact, that a study by ZK Research found that 30% of network engineers involved with SD-WANs spend at least one full working day every week trying to find the source of SD-WAN problems and finding ways to fix them.

There is a possible solution to this problem, and one which is likely to become increasingly common as SD-WAN configurations get ever more complex in the future. The solution is to harness the power of AI to reduce the incidence of problems in network configuration and management caused by good old-fashioned human error. In the near future it is likely that this type of AIOps will be baked into SD-WAN solutions right from the implementation stage.

Fewer outages, more services

The result of this, if and when it happens, is that the complexity of SD-WAN technology will be able to increase, with the ultimate result that this type of AIOps won’t just be a good idea — it will be the only way that managing an SD-WAN will be possible. And the benefits could be significant. That’s because organizations that automate 70+% of their network change activities will reduce outages by at least 50% and deliver services 50% faster, according to a Gartner report.

And it does look like AIOps for SD-WAN is happening. VMware acquired AIOps outfit vendor Nyansa in early 2020 and its AI technology is now part of VMware’s VeloCloud SD-WAN platform in the form of VMware Edge Network Intelligence; Masergy has introduced its own AIOps solution for SD-WAN; and HPE’s Aruba Networks has also added AIOps to its SD-Branch product which includes SD-WAN.

Operational Interdependence

So, as we have seen, IoT, SD-WAN, and AI have a high degree of interdependence. IoT is set to grow, but that will be hard without SD-WAN technology. SD-WANs will have to become ever more complex to deal with the growth of IoT, but that may only be possible if there’s an increased use of AI. And as for AI, it may be that in AIOps for SD-WANs that the technology finds a truly killer application.

Read next: The Impact of Networking Evolutions in 2021