Thank AI for Your Data Gravity Woes

Data gravity is not a good thing. But in today’s data-driven, AI-crazed landscape, it is fast becoming a reality that impacts everyone.

The concept was put forward by Dave McCrory, currently in Digital Realty, in a 2010 blog. It is simple and alluring.

Gravity attracts, and all bodies of mass have this force. A larger mass will have a greater force of attraction and overwhelm smaller objects and their own gravitational pulls. It keeps the planets in orbit and helps with their creation.

In data science, data gravity creates an immovable force to reckon with. McCrory theorized that as companies amass data in one location (a central data center), it will inadvertently influence application and data decisions.

Like gravity, data gravity is inevitable in today’s data-driven landscape. Unlike it, it is something to be avoided.

Data gravity can also be self-perpetuating—as the data mass grows bigger in one location, it becomes more challenging to move data to another location. Over time, this creates latency headaches.

Data gravity is becoming real

The sobering truth about data gravity is that you can’t avoid it. When McCory came out with this observation, cloud adoption was beginning to pick up pace. He claimed that latency and throughput would act as “accelerators”.

With application speed, productivity and real-time insights becoming vital (especially after the accelerated cloud and SaaS adoption during the COVID-19 years), it is easy to see why we have a global data gravity problem.

The latest Data Gravity Index 2.0 confirms this global headache. Daniel Ong, director, solutions architect for Asia Pacific at Digital Realty, notes that enterprises are stuck in data gravity for many reasons. For example, the increase in data regulations and drive toward data sovereignty are seeing companies amassing data in a single geographic location.

Ong notes that the data gravity problem morphs according to your industry. For example, it is easy to see why finance enterprises will create data gravity in New York, Japan, Singapore and Hong Kong.

Another primary driver is customer and employee experience. “If you're an enterprise customer using cloud services, servers, compute, storage, and other services, there will definitely be a concern about where you place your data. It becomes a data placement issue,” explains Ong.

AI adds a new dimension

AI adds another layer of complexity, especially with the sudden FOMO or relentless drive to experiment with the generative flavor.

First, enterprises are opening their historic data troves for machine learning. And that’s a lot of data — Ong remarks that 80% of the data still resides with enterprises based on the report.

This may turn your enterprise into a data gravity, attracting more AI applications that use those learned insights. “So when we say data, gravity, we don't actually mean data centers alone; it’s where you store, and it can be your infrastructure,” explains Ong.

Another concern with AI is that where a model is created through machine learning and where it is used can be different locations. Besides, some applications, like autonomous vehicles and industry 4.0 robotic applications, require the model and computer power to be closer.

This is something that Ong already sees with his customers building an AI platform or using an NVIDIA cluster system. “You want to be as close as possible to that piece of data,” he adds.

Of course, this adds to cloud cost headaches in the real world. With an interconnected architecture, data egress spends may balloon. “A lot of customers didn’t feel it before; now, they will with a lot of data and data traffic going back and forth,” says Ong.

Solving data gravity needs an ecosystem

The Data Gravity Index 2.0 report revealed insights into how big the data gravity problem is becoming. And it also calls for companies to take the issue seriously.

For example, while data gravity is becoming a global problem, the report identified a correlation between enterprise data creation and data gravity with GDP. Essentially, as GDP rises, so will data gravity issues.

Technology maturity can also increase data gravity “as more combinations of technology are used,” points out the report. Meanwhile, it also noted we cannot trivialize the impact of local regulations or digital workflows driven by M&A.

Meanwhile, cybersecurity, the adoption of hybrid and multicloud, and the increased adoption of predictive and generative AI are adding to the data gravity worldwide.

So, while there is no magic bullet to solve data gravity, the report highlights some ways enterprises can reduce its detrimental impact.

So what can enterprises do?

Ong notes that one way to address data gravity is to calibrate capacity across employee centers, then place data and controls cloud-adjacent.

“So what IT leaders want to do is to place more and more data around the world as close as possible to their customers or their employees. That's the use case that we're seeing a lot of customers deploying in multiple locations,” explains Ong.

“You will also need to create centers of data exchange,” Ong adds, citing the report’s conclusions. In addition, the report notes the need for an increase in incremental servers globally and the need for large numbers of servers required outside yet adjacent to the public cloud.

Equally important is factoring in AI and other data-driven applications upfront and understanding whether you are in a data-intensive industry.

Admittedly, no one enterprise can fully address data gravity on its own. It also needs the involvement of an ecosystem of network providers and hyperscalers.

This is where Digital Realty is looking to play differently from its competitors. It is creating competencies and subject matter experts to address each enterprise’s data gravity needs from a use case perspective.

The company is also creating whole ecosystems and working closely with various technology-building block providers and consultants.

For example, the company is building out hyperconnected data centers called Campus Connect that offer Layer 1 connections. It is also moving with the customer needs by becoming one of the first global partners to become NVIDIA DGX H100-ready for the company’s newest data center in Osaka, Japan, KIX13.

“It’s how we connect the dots. Then [our customers] will see a lot more synergy,” says Ong.

Winston Thomas is the editor-in-chief of CDOTrends and DigitalWorkforceTrends. He’s a singularity believer, a blockchain enthusiast, and believes we already live in a metaverse. You can reach him at [email protected].

Image credit: iStockphoto/Pitris