The IT industry is about to shift the ground under your feet – once again. After a decade of pushing the virtues of cloud computing the technology sector is now looking at its next great adventure – edge computing.
Edge computing recognises that to meet new emerging requirements, for example with Internet of Things applications, processing power needs to be pushed out closer to where tasks need to be executed – hence the edge.
According to VMWare Executive Vice President, Strategy and Corporate Development and General Manager, Telco NFV Group Shekar Ayyar, “While concentrated hyper-scale capacity was good for a certain class of application the next generation of applications driven by IOT and machine learning are inherently going to require a much more distributed capacity.”
He told Which-50, “Edge is the incarnation of how that sort of distribution happens.”
Think of the edge as a kind of virtual boundary that is contextual to the use case. A retail outlet will have its own edge, so will the retailer’s warehouse, as will the truck that drives the goods between them. Likewise, the cloud infrastructure company providing all three with their compute and storage services.
“What is now happening with things like IOT as well as with technologies like machine learning and AI applications, there is a new need for a distribution capacity. I can’t always afford to go all the way back [to a central cloud] to get a particular task done, and then return it all the way back for execution,” he said.
Instead, Ayyar said, the better approach is to break the application problem into multiple steps.
“Some of those steps might need their answer very quickly and I might only need to go as far as you and I are sitting away for each other to get the answer.
“For other steps, I might need to go a little further, and for others, I might need to go all the way back.”
Not just yet
But there’s a problem. While edge computing is to set to enter the general business conversation in the years ahead, the reality is that current architectures are not fit for that purpose today.
We asked Ayyar what needs to change. He identified three issues in particular.
“Architecturally it is almost like going from mainframe to client-server computing (The last big architectural shift before cloud computing) where processing was pushed out from centralised computers like mainframes or Unix servers and out to the user’s desktop (client).
“Today’s model of cloud is largely built on the premise of these hyper scaled clouds that are based at specific points and distributed around the world. But you can’t get a cloud wherever you want it. if you want one in a school you can’t just manufacture it there. So the first problem is we don’t have a distributed cloud capacity that is needed to support edge computing.”
The second problem is both technical and economic. “What is the right form factor for vendors to produce the stuff that goes into these edge devices? (For instance miniaturising devices to fit into cars for autonomous driving).”
Finally, there is a networking issue. “Once you have this kind of distributed computing framework then you need sophisticated networking to identify and connect all of these points with the right other points to then complete these services.”
“All of these are required to have a successful edge architecture,” he said.