As a privately managed backbone, MPLS suffers none of the erraticness of the Internet. Yes, that’s old news, but it has been commonly said  that in the SD-WAN vs. MPLS dilemma, SD-WAN will probably replace. What’s happening, though, is something very different.


Despite SD-WAN adoption, companies continue to invest in MPLS services. They invest not out of love for MPLS, but because Internet-based SD-WANs do not provide the necessary quality for running all critical applications.


Only by replacing MPLS with a suitable alternative service can companies realize the full benefit from their SD-WANs.



SD-WAN’s Biggest Problem


It’s no secret that every IT team commits to meeting a stated or assumed level of quality to the business. Perhaps it’s formalized in a service level agreement (SLA) or maybe it’s just a tacit understanding of “goodness” for a given application.  Regardless, an expectation around the quality of experience with an application is always present. When it’s not met, user frustration mounts and support calls increase.


Meeting those quality expectations with most applications is straightforward. As long as applications are tolerant of high amounts of or fluctuations in latency and packet loss, Internet-based SD-WANs are a fine solution. Backup, data replication, web browsing, and email (usually) are examples that come to mind.


But there’s another class of applications that are less tolerant of the random fluctuations in latency and packet loss that are common to the Internet. Voice is the classic example, but so can be many business-critical, custom, or client-server applications. Other examples may be use case dependent, such as large file transfers across long distances.


Running these applications across the Internet may meet user expectations most of the time, but not always. And when application quality suffers, IT is put on the defensive with users and senior management.


The only answer has been for companies to retain their MPLS services to carry traffic from these high value, critical applications. Having the SLAs provided by MPLS operators around availability, latency, and packet loss instill confidence and plays well with the CFO.



MPLS Headaches


But retaining MPLS only makes a bad situation worse. MPLS services are extremely expensive when compared to the Internet. MPLS bandwidth can cost  between three and ten-times more per megabit than Internet bandwidth.


The high cost of MPLS bandwidth impacts more than just budgets. Because MPLS is so expensive, IT managers must be judicious with their bandwidth spend. Branch offices can often only be sized with just large-enough incoming bandwidth for today’s needs and moderate growth. This becomes a problem as traffic loads increase, which is what’s been happening as we adopt more video and other large data objects.


Agility is a problem for MPLS services, and that applies to new lines and line upgrades. Increasing branch office capacity means days, and even weeks, of waiting for the carrier to reprovision an MPLS line. And new site installations can take as much as six months to deploy.


What’s more, private clouds, SaaS applications, and the Internet - today’s destination for most enterprise traffic - often underperforms across many MPLS-based backbones. Internet-bound traffic must be first sent to the sole MPLS-connected site with secured Internet access. The result is the well-documented trombone effect.


Finally, no MPLS service provider has a network that reaches every location. Invariably, service providers need to partner with other vendors to reach the other one or two offices. Once that happens, troubleshooting and diagnostics becomes more difficult and costs increase.



MPLS-like Without MPLS Pain


MPLS services might be out-of-step with today’s market, but the need for a global, predictable, inexpensive service is very real. Businesses have only been able to reference MPLS because that’s what’s available, but there is another possibility.


For years, ISPs reached the greater Internet by paying for capacity across Tier-1 IP backbones. IP transit, as this called, is more expensive than exchanging routes (peering) with other providers. In return, ISPs gain MPLS-like guarantees around availability, latency, and packet loss at a fraction of MPLS’ costs.


IP transit services alone cannot answer all of the deficiencies of MPLS. There are still challenges when it comes advanced security, service delivery, and geographic reach. But Internet transit services are an excellent start towards delivering a predictable and affordable backbone for today’s businesses.


By finding service providers whose business-oriented data services are built on IP transit, enterprises can reap the cost and capacity benefits of the Internet without the headaches of MPLS. And in the process, SD-WANs can become the affordable and effective WANs they’ve always claimed to be.



Dave is the secure networking evangelist in the office of the CTO for Cato Networks with more than 20 years of experience in IT and telecoms. Cato does not provide MPLS services.