How to Intelligently Manage the Last Mile of SD-WAN

Note: This article describes how to intelligently manage the last mile of SD-WAN from the perspective of network architect Matt Conran.

Providing a global SD-WAN is very different from an on-premises network. An on-premises network provides complete control for an end-to-end design, enabling low latency and predictable connectivity. There may still be power outages and brownouts, but you are in control and can troubleshoot accordingly with proper visibility.

Therefore, with a global SD-WAN, managing the middle mile and the last mile is more challenging. Most SD-WAN vendors cannot control these two parts, and if these two parts are not done well, it will affect application performance and service flexibility.

[[250904]]

In particular, SD-WAN equipment vendors often overlook the issue of managing the last mile. Previously in the era of Multiprotocol Label Switching (MPLS), providers bore the responsibility of managing the last mile, but this is no longer the case with SD-WAN. For many global SD-WANs, managing the last mile is a challenge.

Managing the middle mile

As you know, an Internet connection consists of three parts: from the customer to the ISP (the first mile), the middle mile (called the Internet core), and the last mile from the destination ISP to the customer.

The middle mile includes autonomous and interconnected networks, all of which have different business objectives. In the economics world, traffic is coupled to financial exchanges and ISP routing packets. Unfortunately, other metrics that may be more appropriate for application performance are not considered. This can lead to unreliable end-to-end connectivity. On one day, your internet traffic may have few hops, but the next day it may bounce around the globe. There is nothing you can do about this situation. A more feasible solution is to privatize the middle mile. This solution provides strict requirements for latency, loss, and jitter.

A global independent backbone helps solve these problems. Cloud-based providers, such as Aryaka and Cato Networks, offer global backbones. In addition, Mode offers a global backbone that works with any third-party SD-WAN device.

Managing the last mile

In a previous consultancy job, I had to launch a global network in a very short time. There were connectivity issues and I had to request an emergency ticket from the headquarters (an ISP based in Hong Kong).

The CEO had a tight deadline and wanted the ticketing system closed the same evening, not caring about the details. I was at the headquarters in London, and there was a time difference between Hong Kong and me (Hong Kong was 8 hours ahead). It was a long night for me.

Whatever my frustrations, at least I know that the SD-WAN equipment vendor is responsible for keeping the last mile connected. To some extent, I can help, but fundamentally troubleshooting is not my responsibility.

More Websites Mean More ISPs

Each site requires an ISP, and managing the last mile can be a headache, even for the most patient of people. Each ISP requires new relationships, languages, cultures, and processes.

Depending on the size and location of the ISP, I may be assigned a person to walk me through the process and use that person as a point of contact for future questions. An effective network engineer needs to build process shortcuts for each ISP. But what happens when that person leaves or the process changes?

Problems I see in the last mile

I have yet to see a solution that solves this problem. Carriers and providers that monitor and manage the last mile are often limited by the capabilities of edge devices.

Most of the time, this is a standard layer 3 device that uses some simple metrics like Internet Control Message Protocol (ICMP) request/response. ICMP request/response is very low in the stack. It omits some performance-related information that would help ensure that your application is working efficiently.

Operators and providers lack visibility into the characteristics of a link when something starts to go wrong, such as when a link slows down due to interface congestion. Many last-mile management providers focus only on detecting line faults, which does little to improve link speeds.

Carriers and traditional last-mile management providers like Experio manage the line from the customer router to the ISP, they will not detect problems and lack visibility into the ISP's upstream connection.

Last Mile Management Options

There are several ways to solve this problem.

Aryaka is probably the first independent global trunk service specialist to introduce last mile management. Their service monitors the tunnels that traverse Aryaka’s ownership from the customer location to the Aryaka trunk.

Internet traffic is distributed locally. However, this means that they only manage site-to-site traffic. If the ISP is down or slowing down due to routing or QoS issues, Aryaka will not detect this.

Cato recently announced a new service to overcome this gap. The Cato Intelligent Last-Mile Management (ILMM) service will solve the above problem. Cato manages the last mile from the client to the Cato PoP. Both types of traffic are tunneled back to Cato's PoP and the traffic is separated from there.

The ability to simplify last-mile management is a major advancement in global SD-WAN deployments. For each network design, I try to simplify as much as possible while maintaining adequate control and visibility.

Original link:

https://www.networkworld.com/article/3322505/sd-wan/getting-smarter-about-managing-the-sd-wan-last-mile.html