Vertiv’s Martin Olsen Outlines 4 Kinds of Edge Applications at CAPRE’s Dallas Data Center Summit
DALLAS, TX – The Edge is changing before our eyes, but we may not have seen anything yet, especially when you consider the advent of 5G. That’s why CapRE makes it a priority to understand where the Edge is headed next, and how it will get there. So at our recent Texas Data Center Summit, we organized a round-table about this very topic. And the conversation kicked off with some remarks about how exactly to structure such a big and exciting conversation.
“I recently had the opportunity to read about a structured way of looking at the Edge and today’s topic is Life on The Edge: The Advent of 5G, Micro Data Centers & Evolution of the Next Generation,” mused Moderator Donald Mitchell, Data Center Division Manager at Victaulic. “So in trying to give some structure to that, I really liked an article I read lately by [panelist] Martin [Olsen, Vice President – Global Edge and Integrated Solutions, Vertiv Co], which looks at the Edge as basically four architectural types of data. I’d like to start out with that.”
After some brief introductory remarks, Olsen dove right into those four types, starting out with perhaps the most widely used.
“There’s data-intensive, which is the use cases where it is impractical to move all of this data back and forth constantly,” he began. “A Content Distribution Network is a great example of that. Virtual reality is a great example of that. It would be extremely impractical to try to connect back into the core or the cloud every time you want to watch House of Cards. So you’re looking at more of a hub-and-spoke model.”
“You’ve got to push content out that is being consumed in some fashion,” he stressed. “Typically for data-intensive type applications, the archetype is that the data linkage is singular. It’s data going out and being consumed out there and there is very little interaction with it. So as you can imagine, the connectivity is not always but certainly frequent.
“Then you have the latency-sensitive,” he listed, using augmented reality as an example. “You walk into a retail store, and it detects your Bluetooth or what not, and in this case, they’re not connecting you to the history of cash like Bitcoin. But they might link you up to their website to look at your shopping experience. Not a lot of content going back and forth now, but latency is critical here. Latency from a human standpoint, which is several milliseconds, and is often the difference between having fun and throwing up, especially when you’re in virtual reality and 3D. So there are different requirements – there’s more on-premise compute going on. There’s some connectivity outbound that’s letting you know you’re going through the store.
The third “bucket” of Edge applications, according to Olsen, is machine-to-machine latency sensitive architecture. “Latency requirements for machines are completely different from humans,” he asserted. “This is when you’re in advanced manufacturing, you’ve probably heard of the industrial internet of things. There are thousands of sensors within a high-tech manufacturing facility, and the latency between those happens between machines.”
“Obviously, it’s very critical and it happens very fast, faster than a human can appreciate,” he stressed. “So that’s a different requirement from a compute standpoint. It’s very insular – so it typically happens in a campus mode, if you will. But connectivity outbound is not necessarily that high. But when it is, it’s typically for machine learning type of data, when you move large amounts of data out of the factory to a place suitable for high performance compute, to analyze that data. So latency is not necessarily outbound-critical.”
“Then the fourth one is hyper-critical,” he concluded. “Those are typically embedded compute types, like autonomous cars. Lots of compute on board. There is an always connected element to it and latency is very critical. And in this case, you don’t necessarily move a lot of data. So bandwidth is not necessarily the highest priority. Latency and connectivity into a backbone to communicate out, to communicate with other cars around you, are.”
He then repeated that these four archetypes drive different types of requirements. ”But it all centers around the data profile, the communication in and out and within, and then the capacity profile – is it hub-and-spoke, is it pushing data out that’s being consumed, is it very discrete, is it more of a campus-style, is it your [final] frontier where it’s embedded compute and there are harsh environments?” he rhetorically asked the room of 300+ attendees, preparing to wrap up his remarks.
“As you can tell, wireless and 5G plays a huge factor in some of these, certainly compute and what you need to get out there,” intimated Olsen. “Some of them are more storage-intensive with less power density, and some of them are more compute-intensive with higher power density.”
For more from Olsen, check out a previous CapRE Insider Report: Vertiv’s Martin Olsen Simplifies How to Think About the Edge