If you’ve ever been to a Ripley’s Believe it or Not Museum, you’ll know it’s filled with things that are true, but hard to believe. For example, the museum has “ginormous” bugs that you can keep as pets, use to fight one another, or even eat! Hard to believe, but true. They also have microscopic sculptures of things like the Statue of Liberty and Eiffel Tower in the eye of needle. I know it’s hard to believe, but it’s true. Ripley’s also has actual sword swallowers on display. I’ve always found this hard to believe, but these people are actually true.
One thing not yet in the Believe it or Not museums, but should be are organization that use Software Defined WANs that leverage Internet connections to pass mission-critical traffic. I know this type of network defies logic, is as mind boggling as the mysteries of the universe, and takes a faith akin to what Dr. Gaius Baltar had at the end of the Battlestar Galactica series… but it’s true, you can build a network using the Internet. Hard to believe, but true!
Using the Internet to Replace MPLS
Earlier this month, I did a webinar titled, “How to Build a Successful SD-WAN Architecture” with Damon Ennis, SVP of Products at Silver Peak. In the webinar we defined what a SD-WAN is, provided some tips on building an SD-WAN, and explained why the Internet should be considered a viable alternative for MPLS. Organizations that don’t want to jump in with both feet should at least consider offloading some of the traffic onto an Internet pipe.
During the webinar, both Damon and I talked about how the Internet could be used as an MPLS replacement. One of the points we both made was that while a single Internet connection couldn’t not provide the same level of performance of a dedicated MPLS connection, multiple Internet connections that use technology such as path selection, WAN optimization, and QoS could indeed meet or exceed MPLS performance levels.
During the question and answer section, it became clear that, despite both Damon and I endorsing the strategy, the concept was still hard to believe for many of the attendees.
Here is a sampling of some of the questions and comments:
- “Does the Silver Peak solution (Unity) help to guarantee bandwidth of critical applications?”
- “What about latency-dependent applications? On the Internet we can’t control latency, on MPLS we can”
- “We currently run voice/video across our MPLS — would this kind of workload be supported over SD-WAN? What of jitter and other issues that affect voice?”
Based on the large percentage of questions that focused on application performance running over Internet links, it seems that there is certainly a high level of interest in making this shift. but there is also a large amount of skepticism that it works.
For anyone reading this who isn’t a believer, I’ll reiterate the advice we gave on the webinar: make this shift in steps. Things that are hard to believe not only need to be proven, but proven over a long period of time.
The first step should be to shift to a hybrid WAN where Internet connections are used for non-mission critical traffic such as e-mails and remote backups. Keep your VoIP and video traffic on the expensive MPLS connections. If nothing else, this will free up bandwidth on the MPLS links and should enable organizations to defer any kind of upgrade.
Once the Internet has shown to be reliable, try migrating a few small branches to multiple Internet connections. Two or three Internet connections will still be much lower in price than MPLS. Multiple Internet links can provide “five-nines” of reliability if the proper technology is used to build the network. This would be — at a minimum — path selection and WAN optimization. Ideally, the shift to a software-defined WAN would happen at the same time and traffic would be directed via an orchestration tool.
On the webinar Damon discussed how Silver Peak’s Unity Fabric could be deployed on the network to build a software defined WAN. Unity Fabric can be used in a hybrid environment (MPLS plus Internet) or with pure Internet-based WAN giving businesses the option to choose the architecture they are most comfortable with.
Once there is comfort that the Internet WAN does indeed work for the few branches it supports, then migrate over to a full Internet WAN.
I certainly don’t believe that every organization will take this bold step, but those that do will have a network that likely performs as well as an MPLS network but at a much lower cost. All those things in the Ripley’s Museum are hard to believe, but true. Perhaps it’s time for network managers to suspend skepticism and try the “hard to believe”… you just may be surprised.
Image credit: Elliot Brown (flickr) / CC-BY