This content is part of the Essential Guide: A comprehensive guide to enterprise IoT project success

Edge network key to IoT data collection and transmission

IoT might be the next big thing, but the next big hurdle for CIOs will be figuring out ways to architect for it.

CIOs planning to provide the systems -- or form the right partnerships -- for their companies' Internet of Things (IoT) projects might want to take a few pointers from Kris Alexander, chief strategist of gaming and connected devices at Akamai Technologies Inc. The Cambridge, Mass.-based cloud services provider, which accounts for between 15% and 30% of all web traffic, is well known for delivering large content files to consumers. But as connected devices proliferate, Akamai is becoming a force in its customers' IoT projects.

Last May, Akamai provided complimentary service to IoT and wearable startups at the Wearable World Congress. And last month at a Massachusetts Technology Leadership Council IoT event, Alexander said its customers are increasingly turning to Akamai for IoT help. The potential business opportunity is a logical one: Customers such as automobile manufacturers might use Akamai to distribute software updates to vehicles. Now, they're asking Akamai to provide services in the opposite direction -- collecting sensor data emitted by cars for proactive maintenance and service, he said.

The problem for Akamai -- and for CIOs trying to wrap their arms around IoT -- is that collecting and transmitting IoT data isn't as simple as flipping a switch to reverse how data is collected and transmitted, Alexander said. For one, not every IoT data point should be treated equally. Instead, businesses need to pool and prioritize data at the edge -- a design principle Alexander describes as "scaling in."

Collecting IoT data at the edge

When Akamai pushes content out, it relies on an edge network. The "edge" for Akamai, is "an edge of compute close to the user -- within the local [Internet Service Providers] and even out to the home or the enterprise -- to create levels of amplification," Alexander said. Part of Akamai's edge network is made up of more than 100,000 servers dotted around the world where large files can be cached, replicated and made available for streaming in closer proximity to the viewer.

With IoT data, the transmission problem is inverted, Alexander said. He described IoT data as small, frequent and plentiful (Gartner predicts there will be 25 billion connected devices by 2020). Rather than amplify out large content files, such as software upgrades, CIOs will need to funnel the bits and bytes of IoT data in. But transmitting data in a one-to-one ratio from the device to a data center can be inefficient and create chokepoints in the network. IoT data, in other words, needs to be processed en route. "A lot of architectures and networks haven't been built to do that," Alexander said during a panel discussion.

Kris AlexanderKris Alexander

That's where the edge network comes in handy. Akamai is now using its edge as an intermediary location to collect, thin and even normalize IoT data before moving it on to a centralized data center. Pooling the data at the edge first helps Akamai -- and its customers -- avoid network congestion and latency. "If you have millions of people connecting into the enterprise, each doing this one-to-one data feed to the data center, it doesn't scale," Alexander said.

Besides, transmitting every bit and byte of IoT data to the data center is often unnecessary. "If you're looking at a sensor or a switch and the switch is on, you don't need to be getting a status every second in writing to the data center that it's on," he said. "You want to write [to the data center] when it turns off. That's the event that matters."

Transmitting near real-time vs. historical IoT data

While edge networks help structure how IoT data is transmitted, they also pose challenges. One of the biggest? Knowing what questions need immediate answers. 

"There may be data you save for additional analysis later," Alexander said, "but you want to know what questions you need to answer more immediately so that you can get answers in a timely fashion."

Akamai not only thins and normalizes the data, it also uses its edge network to "tier" the data -- all processes that are automated. Data needed for real-time decision making is given a priority status over data that isn't. Simply put, Alexander said, Akamai uses different pipelines to connect the edge to the data center: One pipeline, a superhighway, is reserved for a small set of near real-time data needed to make a quick business decisions. Another pipeline is reserved for a larger data set that businesses use for historical purposes.

Even data slated for transmission via the second pipeline is processed to some degree at the edge. Alexander gave the example of an entertainment company that may be interested in knowing how many viewers watch a program in HD. "Do you need that line of data for each individual user? Or do you need a line of data that represents the population of users?"

Knowing what questions to ask means CIOs can avoid bogging down the network as well as bogging down their analytics technology with unnecessary IoT data. That means businesses can get to the answers faster, an important factor when making decisions in near real-time.

Let us know what you think of the story; email Nicole Laskowski, senior news writer, or find her on Twitter @TT_Nicole.

Next Steps

The IoT is spurring a new architecture for relaying data, referred to by same as mesh networks

Concerns about quality IoT, eased by edge computing

The pros and cons of adding edge computing to a cloud architecture

Dig Deeper on Internet of things (IoT) for business