Last month I tackled the idea of WAN optimization for desktop virtualization and the many challenges that users face when their desktop is moved outside of their office walls. Admittedly, virtual desktops that are centrally hosted off-site might be a stretch for many IT organizations, but the WAN and networking challenges that they will have to deal with for any off-premises virtualized solution are still applicable. In a nutshell, when you move something outside of the data center, you have to deal with the WAN. Nowhere does the WAN play a stronger role than in off-premises cloud computing.

By design, off-premises clouds are at the mercy of the WAN because they reside outside the walls of your data center and off the LAN. But also by design, and one of the major differentiators between cloud deployments and traditional hosted environments, is in most cloud architectures the off-premises infrastructure needs to be tied to your on-premises infrastructure; there needs to be some connectivity over the WAN between your data center and the cloud provider. Not only do you have to deal with optimizing and managing data to and from your cloud provider (and users), but now you have to bring those issues in-house. I wish we had a term for this type of infrastructure that includes a WAN connectivity requirement, something like exostructure.

A great example of how applications rely on exostructure can be seen with applications that share messaging and/or metadata traffic with another application in the cloud. On the LAN – the networking infrastructure – the apps may be fine passing messaging traffic across a message bus with a sub-10 millisecond latency. When those applications are spread across a cloud, however, and that latency over the WAN – the networking exostructure – extends to 80 or 100 seconds, those applications may crumble.

As a user, I see this all the time with websites that request “Like” buttons from various external social media sites. It’s not uncommon for many of the news sites I read to get hung-up while loading in my browser because the sites are trying to make an external call to a social media property outside the scope of the application over the WAN. The apps don’t know any better: they make calls out, they get data in, they move through the queue.

The same could be said for enterprise users as well: We don’t know any better when an application moves in the cloud. To us, our CRM is accessed via a URL; it lives in our browser. But we certainly know when the application slows down, and we know on Friday it took one second to load a page but on Monday morning it jumped to four seconds to load that same page, and we call IT. These examples highlight the challenge of managing two new networks as part of the application exostructure: the WAN between the cloud provider and your data center, and the virtualized LAN that is part of the cloud infrastructure – two elements that are typically outside IT’s control.

As enterprise IT begins to embrace the ideas behind cloud computing – dynamic scale, pay-as-you-go, right-sizing – these details are what can define a successful move to the cloud vs. one that fails, possibly very publicly.

Whatever you call the exostructure, it is a critical part of any off-premises cloud solution and impacts two major WAN-dependent components of apps in the cloud: pushing and pulling massive amounts of application data to your cloud provider, and managing and control of the user experience for applications in the cloud environment. As explained above, both of these challenges are paramount to end users. Their user experience is based on how well they can exchange data with the application and how well that application can exchange data between various components. Together, those two issues basically make up your application-level SLAs for the cloud. An app can go down for many reasons that have nothing to do with the actual application going offline, including latency introduced by the WAN. Having total transparency into all parts of the application and network in a cloud deployment should be a requirement for any enterprise IT department.

Although the components of the exostructure aren’t new – we’ve been dealing with the WAN for longer than I’ve been in IT, and the LAN components of a virtual network have been with us for nearly as long – how they work together as one unit and the requirements that we need to place on them for cloud computing are new. Cloud SLAs are based on user experience, which is based on application response over the WAN, which (among other things) is dictated by the LAN in the cloud provider and possibly the on-premises LAN for enterprise IT users. And all of that is part of the cloud exostructure.