Data Gravity is Shifting the Data Center Network. But in Which Direction?

It’s not just a storage problem. Huge datasets are valuable, but difficult and expensive to move.

“Enterprises have a distribution problem,” Morgan writes. “More users, more locations, more apps, more devices and more data equals big challenges. Moving data will become cost-prohibitive and time-consuming. Enterprises will need to move elements of IT away from their headquarters toward where the data is generated and utilized.”

“It’s going to be harder and harder to move (data) around, even though we think about it as liquid,” said Morgan, who noted that it could costs millions to move an exabytes of data over a 100 gbps network.

Research firm IDC projects that the “Global Datasphere” will grow from 33 Zettabytes (ZB) in 2018 to 175 ZB by 2025. Today, more than 5 billion consumers interact with data every day – by 2025, that number will be 6 billion, or 75% of the world’s population. By 2025, IDC says, each connected person will have at least one data interaction every 18 seconds.

“People, software and machines are creating and consuming data faster and in all the places where we work, play, and live,” said Rick Villars, Research Vice President, Datacenter & Cloud, IDC. “The significant increase in data created, aggregated and analyzed in these new locations is contributing to a major shift away from deploying IT in traditional corporate data centers.”

Digital Realty: Megacity Ecosystems

A new study released Monday by Digital Realty digs deeper into these trends, examining the growth potential for four emerging technologies – AI, the Internet of Things,
IoT, Blockchain and 5G. The Global Digital Capitals Index projects that by 2029 these technologies will drive $721 billion in revenue per year to 60 of the largest cities around the world. The research estimates business impact based on metrics including new jobs, new companies, new industries and improved public services.

“The speed of technology innovation continues to accelerate at an alarming rate, creating tidal waves of data, and every company has specific nuances and risks exploited by data gravity that can’t be addressed with a one-size-fits-all cloud concept,” Bishop noted in a recent column for Data Center Frontier. “IT leaders will look to lean more heavily on the experience of data center providers who lead in planning centers of data exchange zones—achieving a proper mix of on-premises and cloud-resident workloads—to solve data gravity challenges.”

Data gravity is shaping up as a global network design challenge. The Digital Capitals report projects that much of the growth will focus on the world’s largest metropolises.

“It’s no surprise that cities like New York, Tokyo and London occupy a lofty position in the report,” said Sharp. “Not only are they hubs for the highest density of digital commercial activity and digitally-skilled workforces, but also their laser focus on the creation of new digital technologies and applications makes them prime examples of what’s possible in the world’s data economy.”

The report’s most striking predictions involve data growth in the Asia Pacific region, which is likely to see blockbuster digital growth from AI, IoT, 5G and blockchain. This chart reflects the cities that are expected to see the largest growth in the next 10 years. All are in Asia or the Middle East.

That’s why Digital Realty has spent the last several years focused on a global expansion of its data center footprint. The company has acquired Brazil’s Ascenty for $1.7 billion (with Brookfield), created a $1 billion joint venture with Mitsubishi in Japan, and last month agreed to acquire Interxion for $8.4 billion, adding more than 50 data centers across Europe.

“It’s about powering your digital ambitions with the right solutions in the right places, everywhere you want to be,” said Bill Stein, CEO of Digital Realty.

As we noted in our coverage of the PlatformDIGITAL announcement, the strategy reflects Digital Realty’s ambitions to grow its colocation and interconnection business, while supporting metro connectivity hubs with wholesale space via connected campuses.

CyrusOne: Rural Cloud Access Points

CyrusOne has acquired 20 acres of land for a data center campus in Council Bluffs, Iowa, which is also home to one of the largest cloud campuses for Google.

“We will build an enterprise data center in close proximity connected with direct fiber path to a leading hyperscalers compute and storage nodes, which we used for hybrid cloud deployments,”  Wojtaszek said on the company’s earnings call. “It will drive significant improvements and performance cost and security, making it a very attractive value proposition for enterprises, particularly those with massive datasets.”

his Microsoft storage unit has 96 bays for hard disk drives in a JBOD (Just A Bunch of Disks) configuration. It can use 14TB drives for a max capacity of 1.2 petabytes in a 4U chassis. (Photo: Rich Miller)

Wojtaszek says there is a “a noticeable change in data center network topology underway, which is being driven by the growing importance of compute and storage nodes.” He said the Council Bluffs project allows CyrusOne “to provide the provide the private cloud through an architecture that a significantly more efficient than the traditional on ramp compute node network topology.”

That marks a shift in the primacy of interconnection, he says.

“In the early days of the Internet, eyeball acquisition was the primary motivator for companies that were looking to distribute their product to customers over the Internet, which in turn placed a high value on establishing interconnected data centers,” said Wojtaszek. “However, what we are seeing now is that the network nodes that dominated the growth of the data center industry for the first 30 years are becoming less relevant and we believe will be subsumed at the hyperscalers compute and storage nodes, which will be where the next wave of Internet value creation comes from. This is something that I think is going to continue to expand over time as the datasets get larger and larger.”

The trend driving this new topology is AI workloads, which often work best with close proximity between storage and compute, but don’t necessarily need to connect to a large number of networks. “This is a trend that has just started and we believe will only accelerate over the next five years.”

“The network nodes that dominated the growth of the data center industry for the first 30 years are becoming less relevant and we believe will be subsumed at the hyperscalers compute and storage nodes.”
CyrusOne CEO Gary Wojtaszek

Digital Realty executives are also keenly focused on data gravity, but say that customer eyeballs also have gravity, as reflected by cloud computing providers deploying more server farms in the suburbs in recent years.

“If you look at where major cloud destinations are, it’s places like Northern Virginia and New York and Dallas and Silicon Valley,” said Digital Realty’s Bishop. “(Cloud campuses) are all supply-side, not demand side. Data stores have to be located adjacent to clouds – but not in Iowa. Iowa is a low-cost hedge, not the edge.”

Wojtaszek says close proximity to cloud campuses and network backbones offers an alternate approach. He noted the huge concentration of cloud capacity along the 41st parallel, along high-capacity East-West fiber routes that track railroad rights of way.  A number of major cloud campuses have been built along the 41st parallel, which runs through Ohio, Iowa, Wyoming and Utah.

CyrusOne says it is deploying 4.5 megawatts of data center capacity in the first phase of its Iowa project, with half the capacity pre-sold. The facility can support up to 24 megawatts. The company may pursue similar deployments in Texas.

“We’ve been working on a number of different initiatives in Houston for our oil and gas customers,” said Wojtaszek. “We’re seeing is that the data sets that they’re dealing with are just so massive that it’s not really efficient for our customers there to put that data into a cloud, because the ingress and egress fees going moving that data from back and forth to the cloud is really prohibitively expensive.

“We’re expecting this type of topology to happen more and more as datasets get larger over time.”