Jan 10, 2017
When I’m talking with customers, I hear about a lot of challenges their businesses experience with their legacy WANs.
In particular, the network architects and administrators talk about the problems their WANs present as applications move from the data center to the cloud.
The challenges occur whether it is SaaS replacing traditional applications, or specific applications migrating to an IaaS service.
The benefits of cloud-based applications are often readily apparent to end users, and to most IT folks. But many networking staff see the cloud opportunity differently. They are the ones stuck trying to make the old WAN do the new tricks required to support cloud applications.
Inevitably our conversation turns to a brighter subject: what’s next for the WAN.
We talk about how a new WAN should operate, and how the enterprise can liberate itself from the legacy constraints. It is becoming clear a new kind of WAN is required: an SD-WAN that is automated and optimized to do the job of a cloud-driven enterprise—and looking forward, that is not just ‘software defined,’ but ‘self driving.’
The job of a WAN is simple to describe: to connect users to applications, wherever they reside—whether that be in the data center, another location or the cloud.
That’s it: one sentence defines the WAN’s job. So why is it so hard for the legacy WAN to do?
Because the legacy WAN was designed assuming that most of its traffic—and its most important traffic—was internal. The WAN’s traffic was branch to data center, or branch to branch. But today branch-to-cloud traffic is the fastest growing traffic—and the most important traffic for many enterprises.
The legacy WAN often forces branch-to-cloud traffic to take trombone routes through the data center, raising WAN costs and impeding applications performance.
Today’s legacy WAN has inherited decades of complexity and an alphabet soup of network protocols (and an alphabet soup of certifications, a topic for another day). Even simple-sounding things—like leveraging multiple WAN links simultaneously, or routing traffic based on quality measurements—are extremely complicated, and in some cases next to impossible.
IT staff spend an astonishing number of hours configuring, deploying and troubleshooting the legacy WAN, relegating key talent to ‘keeping the lights on’ rather than moving the business forward. A legacy WAN impedes IT staff performance.
In many ways, it all goes back to the beginning of the internet, which was designed to survive multiple nuclear strikes by being a fully distributed architecture with no central control.
Obviously the design did many things right, otherwise the internet would not be where it is today. But somewhere along the way enterprises also got stuck with the same design assumption—that every WAN should run like a mini-Internet, with all the same complexity.
The software-defined networking movement challenges the paradigm that you must distribute everything.
In fact, for many network designs—including building an enterprise WAN—there are many benefits to central control or orchestration. A centrally orchestrated SD-WAN can end the hassles of keeping hundreds of sites configured consistently. Customers can improve security and performance. And reduce complexity and costs.
In essence, the centrally orchestrated SD-WAN liberates enterprises from the heavy baggage of complexity, patchwork of technical band-aids, and arcane CLI’s accumulated by legacy WANs over the last 30 to 40 years.
Customers with legacy WANs are now saying ‘enough is enough’. We don’t need more band-aids. We need a WAN that is optimized for the cloud, that can automatically translate high-level business intent into consistent network-wide behavior.
In my next few blogs I will explore how the SD-WAN builds the foundation to increase business agility and reduce cost and complexity. And how the ‘SD’ in SD-WAN evolves from software defined to self driving.
Meanwhile, I invite you to share this blog with any of your colleagues who are still relying on a legacy WAN, or who want to learn more about SD-WAN technology.