Dec 2, 2015
A few weeks ago I presented on the state of Software-Defined Networks (SDN) at this years Open Networking User Group (ONUG) conference. During the section on SD-WAN I made a comment that I thought that hybrid WAN was an unnecessary step on the journey towards a broadband WAN. It appears I touched a nerve on that topic — the minute I stepped off the podium I was inundated with people asking to me to explain myself further — so I thought I would give more details.
I understand many vendors have built marketing strategies around the concept of a hybrid WAN so I’m sensitive to the fact my statement might have caused some ruffled feathers but I do stand by what I say. Much of the value of a broadband WAN — that is, a WAN built on multiple broadband connections — is the significant cost savings over an MPLS-based WAN, or a hybrid WAN that’s part MPLS and part broadband. However, I also believe that if multiple broadband connections are used, and the proper network engineering is done, a broadband WAN can offer performance that is on par with, or even better than, an MPLS WAN or a hybrid WAN.
In my opinion, the value proposition of the hybrid WAN is harder to understand. If the organization is already running an MPLS network then there’s a certain amount of traffic engineering built into the network through the classes of service. Businesses categorize the different applications and put them into separate groupings to optimize performance. Real-time communications fall into one class of service, mission critical applications in another, best-effort into a third, etc.
In a hybrid environment, what would the business run over the broadband connection? Best-effort applications are already separated from the other applications so there’s no performance impact to VoIP, video or other real time services. There could be a small cost savings benefit by offloading some of the non-critical services to a broadband connection and then cutting the amount of money spent on MPLS but realistically, very few companies actually do that.
In a hybrid configuration, the broadband connection could be set up as an active backup link (instead of passive) but in practicality almost all traffic will continue to be routed over the MPLS link, so in this case the term “hybrid” is being used interchangeably with “backup”.
It’s my belief the biggest value that hybrid WAN brings is comfort. That is, it’s an interim step on the way to a broadband WAN that gives some element of “guaranteed” performance because an MPLS connection is involved. The “guarantees” that most service providers offer with respect to MPLS is laughable at best. It normally requires the enterprise having to prove there’s a problem, and even then there are a number of caveats to the service level agreement (SLA), such as only pertaining to the carrier core and excluding the last mile. Remember, if a service provider is willing to give performance guarantees then they will set the performance levels at a point they know they can meet the majority of the time, so it’s hardly a guarantee.
These are the primary reasons I feel businesses looking to evolve to a broadband WAN should just do it without the interim step of a hybrid WAN. If the leadership at the company is nervous about the implications of replacing MPLS with broadband then wait until there are more best practices and case studies that can be used to justify the shift, but why go through the process of re-architecting the network twice when the end goal is the same?
In summary, I get why hybrid may seem appealing, as it’s a way of dipping a toe in the water without having the associated risk. However, the benefits of this step are minimal compared to the associated extra work involved. So for organizations looking to leverage the benefits of broadband, do so — but take the big leap and get the big benefits.