Software-defined wide-area networking (SD-WAN) is one of the most exciting new technologies of recent years. SD-WAN solutions can increase application resiliency, lower the cost of telecommunications, and even increase application performance. If you are not already using SD-WAN, you probably will be soon.
Unfortunately, SD-WAN systems also come with some significant limitations. Given this, some are looking to the hype surrounding AI, and specifically at the way that AI is being used in the IoT edge, and looking to deploy AI to protect SD-WAN networks.
The Limitations of SD-WAN
SD-WAN is, undoubtedly, a significant step forward in networking. The cost-saving that this new technology provides when combined with the ability to intelligently manage networked devices has led to spiking adoption rates. However, SD-WAN networks suffer from a major drawback – they can be time-consuming to work with.
This time investment comes at every stage of an SD-WAN implementation. These systems can be difficult to set up, and adjusting them to meet ever-changing business needs can also be costly. Errors made along the way add another set of troubleshooting tasks for SD-WAN engineers. These can quickly stack up: one ZK Research study found 30% of engineers spend at least one day a week doing nothing but troubleshooting problems in SD-WAN systems.
These problems are about to get worse. The rise of IoT means that there are around 20.6 billion connected devices already. The time cost of configuring and adapting SD-WAN networks will, therefore, grow exponentially over the next decade. It's no wonder, then, that many engineers are searching for a way to improve automation in them.
SD-WAN and AI
Enter AI. In the context of SD-WAN systems, AI offers several key advantages. First and foremost, AI systems can identify the source of networking issues much more quickly than their human colleagues. The same ZK Research I've already mentioned indicates that 90% of the time taken to fix a problem is spent identifying the source. So automating diagnosis of network issues can allow engineers to spend more time fixing problems, rather than looking for them.
Another major advantage of incorporating AI into SD-WAN systems is that AI can remove human error, which is the largest cause of unplanned network downtime. Ultimately, this will allow network engineers to focus on higher-level tasks, such as adapting and improving networks in line with business priorities, rather than spending the majority of their time in fault mitigation.
At a structural level, the deployment of AI on SD-WAN systems draws on an inherent feature. One of the primary advantages of SD-WAN is that the paradigm brings all routing rules into one place, where administrators can centrally manage them. This makes the deployment of AI systems over SD-WAN relatively straightforward because ML algorithms for fault detection and prediction will already have access to all the data; they need to do their job.
AI could also improve security in SD-WAN systems. AI systems help reduce human error in network configuration and management, and this remains a major source of vulnerability for many organizations. However, advanced AIs can go even further than this. This type of deployment aims to use AI to analyze how certain events impact the network, application performance, and security. These systems can then create intelligent recommendations for any network changes, such as isolating the unauthorized use of SaaS apps.
For network administrators struggling to deploy SD-WAN systems, adding another level of complexity to them might not sound appetizing. However, the reality is that organizations looking to deploy SD-WAN for the first time would do well to investigate the possibility of integrating AI into them at the implementation stage.
There are a few providers who can help with this. Masergy recently introduced AIOps for SD-WAN. This system aims to deliver almost completely autonomous SD-WAN networking and directly integrates AI and ML models into SD-WAN management interfaces. Other vendors are looking to deploy similar systems. Open Systems, another managed service provider, recently bought cloud-based Sqooba to add AIOps to its strong network and security services. And the list goes on. VMware recently acquired AIOps vendor Nyansa and rolled it into its VeloCloud SD-WAN group, a move that gives VMware similar capabilities to Aruba Networks, which initially applied AI to WiFi troubleshooting but is now bringing it to its SD-Branch offering. Cisco is another networking vendor with an AIOps story, although it's trying to apply it network-wide, not just with the WAN.
The Bottom Line
Ultimately, the deployment of AI systems over SD-WAN solutions offers network administrators one huge advantage: saving time looking for (and then fixing) networking faults.
This means that AI deployments of this kind return SD-WAN to its initial principles. SD-WAN, let's not forget, was supposed to save time by centralizing network management. In practice, that aim has often been undermined by the added complexity that the configuration and management of SD-WAN systems entail.
Given that, it's natural that network engineers are turning to AI to automate many time-consuming tasks. Just as COVID-19 has accelerated AI adoption, the increased number of IoT devices that will need to be networked in the next few years means that many network administrators may reach the limit of their abilities and resources. That’s why it makes sense to get help from a new type of colleague: AI.
Related Network Computing articles: