SD-WAN technology has become ubiquitous. However, in reality, operating SD-WAN networks may be more challenging than anticipated due to increased complexity and associated operational data, says Patrick MeLampy, Juniper fellow at Juniper Networks.
First, a refresher on what a software-defined wide area network (SD-WAN) is. There are three main elements that make up an SD-WAN:Â
- Centralized orchestration
- Multiple circuits with high availability and/or automatic failover
- Application-based routing for prioritizing critical traffic
With that in mind, there are also three truths that everyone should know about SD-WANs:Â
- They add complexity, making troubleshooting more difficult
- Public cloud connectivity through an SD-WAN can be very expensive
- Their use of traditional tunnel technology adds additional cost, complexity and performance issues when routing traffic from one location to another.Â
Let’s examine each in greater detail and then take a look at how to avoid the typical issues that can arise when using an SD-WAN.
Complexity
At a baseline, SD-WANs add complexity to network operations. In the past, WANs were built to create a single connection between two locations, which made management and repair much easier. However, these dedicated connections were also more costly. The evolution of SD-WANs grew from the desire to lower expenses and enable traffic to travel over public infrastructure, which was less costly.Â
That said, in an SD-WAN, there are multiple paths and connections that traffic can take on its way to its destination. This means more WAN links, more link types, more network paths, more ISPs and more applications, resulting in more data to sift through when user experiences are impacted. This increase in data, sites, links and other variables makes identifying and solving the root causes of network issues much more complicated. Furthermore, SD-WANs often hide problems or failures, leaving networks vulnerable to subsequent failure.
Further, traditional tunnel-based SD-WAN solutions offer limited visibility into network traffic, making troubleshooting even more difficult.
See More: Closed Loop Verification: A Must-have for Network Automation
Performance Issues
Traditional tunnel-based SD-WANs can also cause performance issues. Tunnels are difficult to set up, manage and maintain, and they add unnecessary overhead to data packets, such as double encryption. They also necessitate handshakes between every network hop, encrypting and decrypting traffic multiple times between endpoints. This introduces traffic latency into the equation, which obviously can impact the overall user experience.Â
Traditional tunnel-based routing also makes it very difficult to prioritize or re-route traffic in the event of an outage or failure; by its very nature, the tunnel-based approach makes failover re-routing a much slower process. For latency-sensitive traffic, such as video conferencing, a few-second delay between traffic failing over to a backup tunnel can cause serious interruptions to the user experience.
Connectivity Expenses
SD-WANs have become a popular choice for organizations looking to connect to public cloud platforms like Amazon Web Services (AWS) and Microsoft Azure. However, using an SD-WAN to connect to the public cloud can be very expensive, as cloud providers like these charge companies based on the amount of data going to and leaving their clouds.Â
Put simply, more bandwidth equals higher costs. By using tunnels to route data, a company is already at a disadvantage. The added overhead that’s tacked on and the increased bandwidth requirements increase the costs of connecting to the cloud. In today’s cloud-first, hybrid working model almost every critical piece of company data needs to be accessible in the cloud. Not only do tunnels make it difficult to prioritize traffic, but they also add unnecessary costs.
Solving the Problem
So, what can organizations do to address these issues with SD-WANs? There are a couple of approaches that have proven successful at offsetting the common increase in expenses and performance issues.Â
One option is to incorporate artificial intelligence (AI) technology into their environment to help manage and optimize the network. AI can be trained to conduct real-time analysis of network traffic, meaning it can identify patterns and anomalies quickly – and often before there are known problems – helping ensure that network technicians can speed resolution time. Imagine how long it can take a technician to pinpoint the reason a wireless network is weaker in certain areas. If the problem is not easily identifiable, then it becomes a matter of trial and error, which could mean physically walking around an office with a laptop and eventually finding out that the third from the last router in the building wasn’t plugged in.Â
With AI, the technician could ask why the network wasn’t as strong in a certain area, and the router could have been quickly identified. The time spent on the discovery and troubleshooting of issues is greatly decreased when AI is incorporated, simplifying SD-WAN management, improving network performance and freeing up network operators’ time to spend on more valuable issues. ​​Also, using machine learning, AI algorithms can identify patterns that are associated with cable failures. These algorithms can then be used to predict when a cable is likely to fail based on its current performance metrics.
See More: Five Reasons Why Your IT Team Needs AIOps
The use of tunnel-free routing is another alternative that avoids the issues inherent in tunnel-driven SD-WAN networks altogether. Unlike tunnels, which add overhead to every packet, tunnel-free routing attaches metadata to only the first packet in a session, significantly reducing overhead and improving application performance. With less overhead and required bandwidth, connecting to the cloud is significantly less expensive than if tunnels were used.
Tunnel-free routing also enables faster failover as critical traffic can be redirected over an alternative path without disrupting the user experience. For example, think of a company holding a virtual user conference. It’s critical to both viewers and presenters that the sessions and speeches be shown without lag time or delay. With tunnels, there is always that possibility. Without tunnels, that traffic can be optimally routed to ensure there are no performance issues and the conference goes off without a hitch. Further, removing tunnels enhances visibility into network traffic, enabling administrators to efficiently monitor end-to-end sessions, track KPIs and better troubleshoot issues. On top of that, tunnel-free routing is faster and easier to deploy. Â
While SD-WANs offer many benefits, they also introduce additional complexities and costs that organizations should be aware of. By implementing tunnel-free routing and leveraging innovative AI solutions, organizations can ensure that their important traffic is not disrupted because of traffic latency or an outage that technicians can’t seem to find the source of.
How are you resolving SD-WAN issues around latency and performance? Share with us on FacebookOpens a new window , TwitterOpens a new window , and LinkedInOpens a new window . We’d love to hear what you have to say!