Edge Computing Players Move Beyond Tower Locations

Mike Dano
2/13/2019
50%
50%

The edge computing industry apparently has learned that locating computing functions at the base of every single cell tower actually is cutting it too close to the edge.

As a result, much of the discussion about the near-term possibilities for edge computing is focusing on computing prospects in a city (like, anywhere in that city) rather than at the base of a specific cell tower.

Greg Pettine, the founder & EVP of business development of EdgeMicro , said the edge computing startup is no longer looking to build edge computing data centers at the base of cell towers, mainly because there isn't enough fiber in those locations to satisfy customers' edge computing requirements. Instead, he said the company is looking at other locations that are closer to major intersections of fiber, with plenty of network routing opportunities.

Similarly, startup Vapor IO also isn't putting its mini edge computing data centers at the base of every single cell tower. CEO Cole Crawford said that it's "not relevant to talk about sitting at the bottom of one tower." Instead, he said the company is talking about "tower-aggregated and connected, not tower-located" scenarios.

"We never actually said that they had to be located at cell towers," Crawford explained, noting that Vapor IO's first edge computing data center in Chicago is running inside a Distributed Antenna System hub with a significant amount of nearby fiber.

"It's not that being at the base of towers doesn't work," he added. "It's that when you go to a market and you're looking for the most optimal physical place you can be, when you're talking about the type of traffic that we expect to be talked about on the edge, you want to go to where you have multiple strands of fiber available to you so that you can spin up a number of different backhaul connections to a bigger, regional data center or enough strands of fiber that would allow us to interconnect our own physical real estate footprint with other Vapor edge modules or other Vapor locations."

Continued Crawford: "So it's not that towers are not the most optimal place, it's just they're a data point in about 104 data points that go into how we do site selection, of which power and fiber are two of the most critical."

Vapor IO's Matthew Trifiro clarified though that the company is "going as close to the tower as physically possible" in most cases, and that some of its sites are at the physical base of a cell tower. But he pointed out that wireless networks are increasingly designed to manage hundreds or potentially thousands of cell radios at an aggregation site, which is where Vapor IO is putting its edge computing functions.

This is the same approach that others are taking. For example, Verizon Communications Inc. (NYSE: VZ)'s Adam Koeppe said that the operator's recent edge computing test was physically located in one of the operator's C-RAN hub sites, not at the base of its 5G transmitter. (See Verizon Gets Serious About Edge Computing.)

Regardless, the rejection of individual tower sites as an ideal edge computing location is noteworthy considering such locations have long been trumpeted by players in the mobile space as well as those in the tower sector. (See Tower Companies Brace for an Edge Computing Bonanza.)

Putting edge computing at the base of cell towers did make sense, though. After all, the whole idea of edge computing is to bring computing functions physically closer to customers -- doing so can dramatically reduce the amount of time (latency) that it takes to conduct computing. Such a design would also mark a dramatic departure from today's computing paradigm, in which computing requests are sent hundreds or sometimes thousands of miles away to massive data centers that are typically located in only a handful of major US metros. But reducing the distance of that computing roundtrip is critical for some cutting-edge applications: For example, an 80 millisecond latency drag on a virtual reality game can make a user sick because what they see in the virtual environment won't immediately synch up with their movements.

And though individual cell tower locations may well end up hosting edge computing capabilities at some point in the distance future (if demand for edge computing explodes), those locations aren't working for most of today's edge computing players. That's partly due to the wireless industry's efforts to aggregate radio management at a central location, and partly due to the nascent status of the edge computing market.

Vapor IO's edge travails
In fact, Vapor IO's setbacks in its ongoing journey through the edge computing landscape are illustrative of not only the type of computing locations that players are eyeing, but also of the overall development of the space.

As CEO Crawford explained, Vapor IO did not reach its goal of building 27 edge computing data centers by the end of 2018.

So how many has the company built so far? Two.

That's because Vapor IO's initial edge computing business plan was funded via an investment from tower company Crown Castle International Corp. (NYSE: CCI) (thus giving Vapor IO access to Crown's tower sites). But that plan fell apart last year, partly because of timing. "I think the market was still trying to find its balance," Crawford said. "We also wanted the market to catch up."


You're invited to attend Light Reading’s Big 5G Event! Formerly the Big Communications Event and 5G North America, Big 5G is where telecom's brightest minds deliver the critical insight needed to piece together the 5G puzzle. We'll see you May 6-8 in Denver -- communications service providers get in free!


As a result, Vapor IO sought more financing for its plans, and in September of 2018 managed to raise a Series C round of funding led by Berkshire Partners, with Crown Castle also participating. (Vapor IO did not disclose the amount of money it raised.) "That changed the way were were going to go to market," Crawford said of the new funding.

Under Vapor IO's new edge computing deployment plans, the company now expects to build computing locations in six metro areas this year -- three in each city -- with the goal of eventually expanding to 30 metro areas. Those initial areas will include cities such as Seattle, Atlanta, Los Angeles and Pittsburgh -- locations that, Crawford said, are outside the handful of metro areas that already host major data centers.

Next page: Federating the edge

(2)  | 
Comment  | 
Print  | 
Newest First  |  Oldest First  |  Threaded View        ADD A COMMENT
daveburstein
50%
50%
daveburstein,
User Rank: Light Beer
2/13/2019 | 6:59:21 PM
Next step in analysis
Mike

Very solid reporting. I've spoken to many of these people the last few days, trying to find a way to think about Edge. Two first thoughts.

Will the Edge Servers be outside the carrier networks, as Vapor expects? Hard sell, unless the existence of the Vapor Edge cloud inspires new entrants. Not impossible, one reason so many want to work with Vapor on this.

Deutsche Telekom is well along building and testing an Edge in Germany, designed for 20-25 ms. Will a Verizon or DT want some upstart to take over a primary part of the network?

------

Where will the Edge servers go? Vapor (and most others) believe it's impractical to go to every cell, including small cells. That would take tens or even hundreds of thousands of robust servers. (Think the ability to do near-instantaneous face recognition in a crowd that's moving.)

I'm calling that deployment, right at the Edge, a Level 1 Edge Cloud. 5G that's deploying today is about 10 ms on the air interface, so that's the starting point. The server/router adds milliseconds, perhaps 4 ms. Level 1 is probably years away as is the 1 ms version of 5G.

A Level 2 cloud is like Vapor - 1-3 hops further back, which adds latency. They've ducked my questions on this, but it adds between 3 and 12 ms to be a couple of hops back. 

A few hops back would require thousands of servers for a country like Germany or the U.S.

A Level 3 cloud would be inside the carrier network but further back. It would only take a few hundred servers to cover a country like Germany. The latency on a 10 ms 5G would be 20-25 ms. That's far more than the 1-4 ms dozens will promise at MWC, but enough for many applications. Think multi=player, highly interactive, Pokemon Go.

 

 
Mike Dano
50%
50%
Mike Dano,
User Rank: Blogger
2/13/2019 | 12:35:43 PM
Updated slightly
I updated this article slightly to note the move to aggregated, or centralized RAN as well as to clarify Vapor IO's edge computing locations.
Featured Video
Flash Poll
Upcoming Live Events
September 17-19, 2019, Dallas, Texas
October 1-2, 2019, New Orleans, Louisiana
October 10, 2019, New York, New York
October 22, 2019, Los Angeles, CA
November 5, 2019, London, England
November 7, 2019, London, UK
December 3, 2019, New York, New York
December 3-5, 2019, Vienna, Austria
March 16-18, 2020, Embassy Suites, Denver, Colorado
May 18-20, 2020, Irving Convention Center, Dallas, TX
All Upcoming Live Events