The Geography of Data

Edge computing limits latency by reducing the distance between processing power and data sources.

We’ve all heard that the three most important factors when buying property are location, location, location. It turns out that time-worn phrase also applies to the computing landscape. Researchers find that the physical location of processing and storage resources directly impact computing efficiency.

This is why analysts and engineers anticipate a rapid rise in edge computing — a network design model in which computing resources are placed at the network’s edge in close proximity to data-collection sources such as mobile devices and IoT sensors. Variously known as cloudlets, micro data centers or fog nodes, these edge resources address some of the challenges created when organizations run increasingly data-heavy workloads in the cloud.

Some suggest that edge computing could eventually replace the cloud model. For example, Thomas Bittman, a vice president and distinguished analyst with Gartner Research, claimed in a recent blog post that “the edge will eat the cloud.”

“The agility of cloud computing is great — but it simply isn’t enough,” Bittman wrote. “Massive centralization, economies of scale, self-service and full automation get us most of the way there — but it doesn’t overcome physics — the weight of data, the speed of light. As people need to interact with their digitally assisted realities in real time, waiting on a data center miles (or many miles) away isn’t going to work.”

Long-Distance Runaround

The cloud model transformed computing by centralizing data processing and storage inside large server farms from which organizations can access applications, services and data across Internet links. It is a proven model that creates significant operational benefits while dramatically reducing spending on on-premises infrastructure.

Cloud service latency — the delay between a client request and a cloud service provider’s response — has always been present, but it hasn’t been a huge deal. For most workloads, these delays have been mostly imperceptible. That is beginning to change, however.

With the rapid growth of mobile computing and the Internet of Things (IoT), millions — perhaps billions — of connected devices will soon be collecting and accessing data at the edge of the network. Transferring that data back and forth to far-flung cloud data centers is creating issues with network congestion and latency.

In a landmark 2013 study, computer scientists at the University of California-San Diego and Google found that applications running on cloud computing systems run up to 20 percent more efficiently when the data they need to access is located nearby. They tested applications running in a warehouse-sized cloud server installation, then compared those results with tests on similar servers running in isolation rather than as part of a cloud.

The researchers found that apps running on the isolated servers ran significantly faster and more efficiently. This is partly because these servers have multiple processors that aren’t being shared across multiple workloads. However, they also found that latency was reduced due to the proximity of data resources. Apps requesting data from remote cloud servers had to wait longer for the data they requested to arrive.

A number of factors contribute to distance-based latency in a cloud environment, including the number of router hops or ground-to-satellite communication hops between an organization and its cloud provider’s data center. Packet delays on virtual servers are another source of latency, as are larger workloads.

Studies have demonstrated that even the subtlest network delays can negatively affect business operations. A Google study found that a 20-millisecond delay can result in a 15-percent increase in web page load times. Slower load times are bad for business. Google found that a half-second delay will cause a 20 percent drop in traffic. Amazon found that delay of one-tenth of a second can cause a 1 percent drop in sales.

Close to the Edge

The edge computing model aims to reduce such delays through the use of small-scale data centers that put processing resources only a single hop away from end-users. “Edge” data centers are cropping up in Tier 2 and Tier 3 markets across the U.S. — cities such as Cleveland, Ohio, Nashville, Tenn., Pittsburg, Pa. and St. Louis, Mo. — bringing dynamic content and cloud services physically closer to customers. Data can be cached locally so that it travels a shorter distance, resulting in improved performance.

While location is a key feature of edge data centers, it isn’t the only one. To be considered an edge data center, the facility must reach at least half of the customers in the market, and serve up at least three-quarters of the dynamic content those customers consume. Otherwise, it’s just a data center in a smaller city that hasn’t moved the Internet edge.

Edge computing preserves bandwidth and reduces network congestion by limiting the flow of data between the data center and the cloud. Single-hop data transfers will be particularly useful for running real-time services and improving mobile data analytics. According to new analysis from Market Research Future, the global market for edge computing is expected to grow at a 35 percent annual rate through 2023, reaching a total valuation of $33.75 billion.

“In our digitally connected world, businesses and consumers have a low threshold for interruptions to service,” said Chris Hanley, senior vice president, Data Center Systems, Schneider Electric. “Edge computing solves real-time data transmission issues by bringing bandwidth-intensive content and latency-sensitive applications close to the users or data sources to ensure reliable connectivity.”