For most enterprises with a significant MPLS-based WAN, introducing software-defined WAN and reevaluating how WAN traffic is directed can result in significant cuts in future spending -- what's known as bend the spending curve downward. Additionally, a move to SD-WAN could even save money now compared to current spending.
The purpose of a WAN is to connect users at a site to resources not at that site. Years ago, that connectivity mostly meant connecting sites to applications and storage provided by corporate data centers. Those same data centers also provided access to the internet.
More recently, though, the shift to SaaS and cloud meant more WAN traffic wasn't about inside-to-inside use of in-house resources. Instead, by using cloud frameworks, WAN traffic moved inside to outside company resources. Furthermore, network traffic has even moved outside to inside as remote workers access any remaining on-premises resources. Now, most WAN traffic touches outside traditional organizational network boundaries.
As network traffic has shifted toward off-premises resources and the number of remote workers has risen sharply, organizations need to ask themselves: Do we still need a WAN? And, if so, for what?
Even if organizations moved workloads too hastily to IaaS and yank them back to on-premises data centers, many end users still work with applications and data running in the cloud. So, direct internet access from each site to a cloud provider's network would probably provide the best performance -- no extra WAN trips necessary. But a WAN is still a good idea for the rest of the enterprise work that happens in data centers -- still as much as 40% on average -- or in other branches.
Organizations that need a WAN should ask themselves: Is MPLS still the best choice for the WAN we need now and into the next decade? MPLS networks still tend to perform better than the internet with respect to packet loss, latency and jitter. But internet services have come a long way in the last decade. So, the answer could come down to cost: Can we save money by dropping MPLS and moving to SD-WAN?
Is MPLS more expensive than SD-WAN?
Whether an organization can save money by moving to SD-WAN is complicated to calculate.
The calculations involve everything from the per-site and overall costs of the SD-WAN service to labor costs for networking staff. However, the cost calculations are dominated by connectivity considerations, such as the following:
- MPLS and internet rates.
- The number of connections at each site.
- The number of WAN sites converted.
However, when using a cost analysis model, the SD-WAN vs. MPLS cost comparison is not an either-or choice. In fact, organizations considering SD-WAN and its associated costs often intend to retain MPLS in their architectures. Usually, in these cases, organizations overlay SD-WAN and beefed-up internet services at each site onto their existing MPLS WAN.
For most organizations, they planned to push cloud-bound traffic onto the SD-WAN using local internet breakout functionality. Keeping such packets off MPLS links would enable existing MPLS bandwidth to serve for years longer or even enable them to be downgraded. Smaller sites might have MPLS removed, and new sites would not have it added unless there was a specific need for it.
Given that information, figuring out when an SD-WAN built on top of MPLS and the internet started saving money was complicated. The biggest determining factor was the MPLS premium. Namely, how much more would it cost to get MPLS at a given bandwidth compared to business internet with service-level agreements at the same bandwidth?
At one point, the MPLS premium was, on average, about 90%. That is, MPLS would cost 190% what business internet cost for the same Mbps of service. And the availability and quality of consumer broadband made it an interesting option for businesses with small branches. Also, it was a lot more inexpensive compared to MPLS on a per-Mbps basis.
Most organizations have modeled a future WAN in which a typical site would have an MPLS link capped at its current speed and two internet links to handle cloud traffic, less-demanding internal traffic and growth in demand. Given this setup -- and assuming they were migrating more than 20 sites -- companies would generally see savings within the first year of the plan, compared to handling growth by increasing MPLS link sizes. Essentially, more sites meant faster savings. Less MPLS meant bigger savings faster.
In addition to spending less, users at those sites would typically see better performance for cloud apps that no longer needed a round trip to the data center. Also, sites would have fewer outages or WAN incidents since they'd have multiple live links and not just one.
How can SD-WAN reduce costs?
Over time, the availability of SD-WAN, improving internet service quality and deepening shifts to cloud and remote work have eroded enterprise commitment to keep MPLS. More and more enterprises see MPLS as optional to their future WAN rather than necessary.
Facing these pressures, MPLS providers have steadily cut the cost of MPLS on a per-Mbps basis compared to internet service, reducing the MPLS premium to about 10% for a large enterprise.
So, can SD-WAN help organizations spend less on MPLS? Yes but with the following stipulations:
- The size threshold for savings is higher now -- well beyond 20 sites.
- The amount saved over time will be smaller.
- The payback may not come until year two for a large deployment.
For organizations that still need a WAN, the "save money" rationale for shifting to SD-WAN is weakened by the shrinking savings over time and the break-even on the investment further out in time. But the cost analysis approach alone doesn't address the improvements that SD-WAN can offer, which increases the importance of building a business case that includes SD-WAN.