We've covered a lot of territory during our first four-part reconnaissance tour across the frontier of the world's data centers for ZDNet Scale, in search of what's being called "the edge." The fact that there is an "edge" at all is evidence that "the cloud" is not a cloud, or at least, not one any longer.
Computing is, and always has been, a mechanism. Physics mandates that it works better in some places than in others. When network engineers first drew clouds to depict certain parts of their networks, it was to mask those areas whose functionality we didn't need to know about.
As distributed computing evolves, there's less and less for us to comfortably ignore.
In the first leg, Vapor IO demonstrated that space does not need to be centralized to be useful. If the most desirable place for a data center connection to be is at the point of minimum latency, then the cell tower should carry a premium. There's not enough space within a facilities shack for a hyperscale data center. But since there are tens of thousands of towers, and connectivity between them is optimum, then that shouldn't matter.
Waypoint #2 showed us that the hyperscale data center is not necessarily threatened by the edge model. In fact, as the host for the orchestrator and the key destination of CDNs such as Limelight Networks that take fast shortcuts around the public internet, they could still be the center of all the action.
But in those regions of the world where even the most hyper of scaled-up data centers is inaccessible, as we saw in Waypoint #3, it is indeed possible to create distributed applications that bridge together hundreds of widely dispersed points of business. What's necessary, as Schneider Electric showed, is a model for physical deployment that gets out of people's way.
On the final leg, however, we left open the biggest question of all: With a new, more subdivided data center terrain, where will enterprises inevitably stage their different classes of workloads? Edge models may have the capability to supersede the cloud models of today, and the first-generation virtualization models that are still hanging on. Someone will inevitably investigate why the edge can't be the center of operations. And by the time the edge takes hold, the reasons why may have all evaporated into history.
Both operators and DevOps professionals are paying more attention to configuration and performance, not less. As a result, the components of their new "cloud-native" applications are actively investigating their own environments. Their objective, for any one point in time, is to locate the edge, and when necessary, move to it.
This changes everything about what an application is. Even the apps on your smartphone perform a number of functions that may soon be offloaded to the server side. Imagine if a less capable phone could function just as well as the expensive one you own now. A lot of mobile device manufacturers that found themselves shut out of the current market, could claw their way back. (Windows Phone, anyone?)
So our figurative map of the territory where the edge is being staked out becomes very important. Figuratively speaking, there's no single "edge" on this map, but every part of it represents someplace where a competitor believes the edge might be. Dell Technologies might be happy if the entire territory were divided into cloud, core, and edge segments. AT&T believes in a "customer edge" and a "service edge," depending on the market being addressed. Vapor IO believes it's found a way to populate both of AT&T's edges with the same, premium quality, miniaturized service. And the hyperscale data centers, which are just now deploying a new wave of distributed technologies, still make a strong case that the center is also the edge.
Once we've mapped out the edge, we will have drawn the map for information technology as a whole. And anyone who hasn't staked out a claim to the edge by that point, may as well sit this next era out.
Journey Further -- From the CBS Interactive Network
- How to get the Kubernetes help you need
- Oracle joins the Kubernetes movement
- These two vendors are most likely to bring Kubernetes containers to the enterprise