An Inside Look at Verizon's Edge Computing Capabilities
Verizon last month announced that it successfully conducted a test of edge computing on its 5G network. The company said it was able to cut latency speeds in half in the test.
The announcement was a signal by the company that it's interested in edge computing and is making progress in the area. But the reality is that Verizon already has a comprehensive view of edge computing, and has already deployed many of the systems and technologies that will be required to run serious, revenue-generating edge computing services.
Here are three key data points that, taken together, indicate Verizon has already made substantial progress toward commercial edge computing services:
We'll get deeper into each of these items, but first, we need to set the stage a little bit. Additionally, it should be noted that it's still early days in this sector of the telecom industry, and as a result, some of the key elements in a full-blown mobile edge computing system -- like a widespread mobile 5G network and advanced network slicing -- remain on the horizon.
Verizon's edge computing story begins with what the company currently calls its "Intelligent Edge Network," or iEN in Verizon parlance. Verizon's iEN runs on an internal cloud network built by Verizon using commercial, off-the-shelf hardware. Inside that cloud, Verizon can run virtualized network functions that are managed by Verizon's OpenStack-powered virtual infrastructure manager.
Adam Koeppe, Verizon’s senior VP for network planning, explained that Verizon's edge computing test last month leveraged its iEN, including its cloud network and virtualized network functions, by installing multi-access edge computing software into the servers in one of the operator's C-RAN hubs in Houston. Alongside that edge computing software, Verizon also installed AI-powered facial recognition software. The edge computing test transmitted video of a crowd over Verizon's 5G network to the computing functions in the operator's C-RAN hub, which then quickly scanned the faces in that crowd for potential matches against a database. The goal of the test was to show that police could potentially use edge computing functions to quickly find someone in a crowd, rather than having to wait for that video to be sent to a geographically distant data center to be analyzed.
15 ms latency on a network slice
Koeppe said the latency speeds obtained through the company's edge computing test in Houston were 15 ms or half of what's available on Verizon's LTE network. While that's nowhere near the 5 ms promised by Verizon's Hans Vestberg during his CES keynote about 5G, Koeppe explained that 5 ms is obtainable today under specific network configurations that weren't used in this particular test. (Interestingly, Koeppe also said that Verizon employed a dedicated slice of its network for the Houston edge computing test -- he said the company manually programmed the slice for the test but, in the future, he said the operator would likely do so automatically and programmatically.)
The edge computing test was successful, Koeppe said. He explained that Verizon already has thousands of C-RAN and SAP sites around the country where it can potentially replicate that same test -- locations that already run the operator's cloud and virtualized network functions. All Verizon has to do is remotely send the right software to the correct C-RAN or SAP computing location.
And that's the key thing about edge computing in general: It's a localized service. The only way to reduce latency speeds is to ensure that computing requests are crunched geographically close to the user and that they are not sent hundreds or thousands of miles away to a data center. Koeppe explained that it's those "hops" -- jumps across switches, interconnections and other network equipment -- that add to network latency. And removing those hops is the key to lowering network latency.
"In the context of the LTE architecture, all the way out you've got a cell site. Then with C-RAN you've got an aggregation point for fiber and baseband processing. And then all of those come back to what's called an SAP location. And they're distributed around the country," Koeppe explained. "And in that path, we've got multiple layers of very strategic locations that can be used for edge compute capabilities. So when you look at the use cases that will emerge with edge compute, whether they're a consumer or business driven, we feel we're in a very good position to then rapidly deploy architecture or network capabilities to meet those use cases because of the fact that we have the intelligent edge network already built."
Indeed, Koeppe added that the operator's 5G network is just a piece of a much bigger puzzle: "It's funny because, you know, we've talked previously about the intelligent edge network. There are multiple parts to that. One of those capabilities is edge compute. And the reason we got so excited about this type of test is because it actually brought two together, you got our 5G radio access network, and you had some edge compute capabilities, brought together within our cloud platform on the intelligent edge network."
This is something Verizon is working on now. For example, a recent job posting for the operator hints at "the migration of existing wireless edge routing functions, such as the Choke Router and the Provider Edge (PE) Router, into the new iEN Multiservice Edge Router being placed in all the SAP locations in 2019."
Next Page: Getting developers at the edge