In 2013 Gartner famously asked, Is MPLS Dead? The short answer is no, not yet. But it is dying. Total bandwidth in the world will grow 29% this year, as it has for the past four. Yet MPLS is only forecast to grow 6.7% through 2028. The data show that MPLS has lost its grip on the market.
The real question is why and, more importantly, what’s next?
Why MPLS is dying
MPLS was designed in the late 1990s for a different era. Networks were simpler. The enterprise connected the data center(s), HQ office workers, and perhaps a few branch offices. Workloads were well-known and traffic was predictable.
Fast forward to 2023, and we see a vastly different landscape. Enterprise networks now span data centers, corporate offices, multiple clouds, edge networks, and thousands of remote workers. And, enterprises are increasingly connecting to partner and customer networks.
Further, traffic has grown exponentially since 1997 (50% CAGR). But traffic has also become unpredictable. Workloads like video and unified communication make traffic much more bursty.
These factors have put tremendous stress on MPLS, creating serious issues in three important areas.
- Lack of agility. MPLS is slow and cumbersome to provision and change. There is no central point of operations. MPLS requires expensive on-premises equipment wherever it is terminated within the enterprise. And, MPLS has no mechanism for automatically scaling to meet traffic demands.
-
Architectural shortcomings. MPLS is only available where it is available. Coverage is quite broad, but often unavailable at remote edge locations, where enterprises are doing some of their most innovative networking.
Also, MPLS doesn’t support advanced functionality, such as rate shaping or load balancing. - Cost. MPLS costs as much as 200 times the price of internet connectivity. This combines high staff skill requirements and expensive point-to-point on-premises infrastructure to make MPLS prohibitively expensive.
With all these issues, it is no surprise that MPLS is dying. What is surprising is that it isn’t dead yet. The reason for that has more to do with the shortcomings of what enterprises have been trying to replace MPLS with.
SD-WAN missed the mark
SD-WAN emerged as the heir apparent to MPLS in 2012. It promised to fix MPLS’s cost and agility problems with a software-designed approach to building networks and its ability to combine MPLS and public internet bandwidth. For a time, SD-WAN showed promise.
But SD-WAN could not keep pace with enterprise networks’ continued growth and complexity. SD-WAN requires the enterprise to set up tunnels between every point. If you double the number of nodes, you quadruple the number of tunnels. And enterprise networks did much more than double between 2012 and 2023.
The overwhelming overhead of setting up and managing these tunnels destroyed any agility gains from SD-WAN’s software-defined elegance. SD-WAN is unmanageable for many complex enterprise networks. In the 2023 State of the Edge survey, enterprises gave SD-WAN a failing grade on agility, scalability, and cost.
If not SDWAN, then what?
As previously stated, MPLS is dying. But until there is a suitable replacement, it will hang on. The question is – what does MPLS’s replacement need to look like? To answer this, it helps return to first principals.
Let’s start with what MPLS got right. That’s simple: MPLS is a private network that delivers predictability and privacy. MPLS delivered guaranteed performance because it didn’t have to worry about the random congestion and outages the public internet faces. And MPLS delivers privacy because, well, it is a private network.
Compare this to using the public internet for broadband, as SD-WAN does in part. Traffic on the public internet has to cross a vast digital wilderness that the enterprise neither owns nor controls. A wilderness into which the enterprise has zero visibility. There is no way to guarantee either performance or privacy over the public internet.
Lesson one – MPLS’s replacement must be based on a private network. That will deliver performance and privacy. But what about agility?
For that, let’s look at what SD-WAN got right. SD-WAN delivered agility through its elegant software-designed architecture. The enterprise used a simple visual tool to design, configure and deploy networks. This is brilliant, and absolutely necessary for agility. Yes, the tunnels required by running over the public networks obliterated those agility gains, but we’ve already shown that MPLS’s replacement must be a private network. The enterprise doesn’t need to set up tunnels for a private network.
Lesson two – MPLS’s replacement must use a centralized visual tool for designing, configuring, and deploying the network. This portal will sit in the cloud, allowing MPLS’s replacement to be delivered as-a-Service.
And, finally, cost. Why is MPLS so expensive? The reason is that MPLS is old technology in which the control and data plane are coupled. This means every part of MPLS infrastructure requires enormous compute capacity to power the control plane functionality.
This isn’t how modern networks work. Today, the control plane is separate from the data plane, meaning most infrastructure requires less compute capacity. The ultimate version of this concept is to have control run in the cloud, where there is unlimited (and extremely affordable) compute capacity.
Lesson three – MPLS’s replacement will separate the control and data planes, placing control in the cloud.
MPLS without the cost and complexity
In the late 90s, MPLS was the right technology for the time. But enterprises struggle to make MPLS work in 2023. A simple new network edge built on three core principles – a private network, a visual cloud-based portal for provisioning and designing the network, and a cloud-based control plane that runs separately from the data plane – will finally allow enterprises to retire MPLS for good.
