Much of the hype around the Internet of Things is centered on a decentralized model of deployment – edge computing, where specialized devices sit close to the endpoints they’re managing or monitoring, is very much the flavor of the month.
Yet the cloud and the data center are still critical parts of the infrastructure, and the huge growth in IoT deployments is having an effect on them, as well. Even deployments that lean heavily on edge compute can stream data back to a central hub for more detailed analysis. So it’s tough to argue that rise of IoT hasn’t changed requirements and expectations in the data center.
What’s much less clear, however, is the precise nature of the changes being wrought in the data center. According to analysts and even the people running the data centers, the jury is very much still out on what, exactly, IoT is doing.
According to Andrew Fray, managing director of European colocation provider Interxion, the one thing that does seem clear is that networking and connectivity capabilities are the main areas in which data centers are being asked to up their game.
“Connectivity is the short answer to the question, but it’s sort of mindful connectivity, depending on what the business is doing and where they want to put the rest,” he said. “So some of that information may need to go to some kind of deep storage, so they may want a very low-cost, high-latency, highly green location. Or they may want a very fast, very high-volume transactional location, in which case [customers' data centers or facilities] probably are going to be close to city centers or positioned within a few miles.”
Architecting for IoT
Serious discussions about architecting for the IoT are just beginning at many companies set to be affected by the new technology, Fray added. This includes which parts of a given workload will live in the public cloud, which will be handled at the edge and which will be worked on by the company’s in-house data center.
“I think we’re increasingly becoming connectivity hubs,” he said. “And what that means for us is that [Interxion]'s finding increasingly connected workloads, and by that I mean people who want volumetrics or they’re looking for [low] latency and proximity. So they’re looking for big bandwidth and/or faster.”
Part of the issue is that the IoT covers a huge range of different functionality and deployment models. Rohit Mehra, IDC vice president of network infrastructure research, said that the implications of supporting IoT as a whole are, consequently, very broad.
“There is a pretty holistic impact on IT infrastructure in the datacenter – including servers, storage, networking, security as well and systems management including APM/NPM and associated analytics,” he said. “Depending on the use cases involved, these could be passive with overwhelming amounts of data going in one direction, versus more active IoT apps that involve automated actions and responses based on the state of the sensors providing data.”
IoT endpoints are even finding their own way into the data center. Smart building systems, like HVAC, temperature sensors and even access control are popping up here and there, according to 451 Research vice president Christian Renaud. But the main way the IoT affects data centers will still be as a capacity driver, particularly for deployments that demand coordination across multiple sites combined with low latency.
“There are applications for this in healthcare and manufacturing, and also pretty much any industry that wants to do video analytics of camera data vs. streaming raw data to the cloud and the cost of bandwidth/ingestion/storage that brings with it,” he said.
IoT impact on data-center management
Beyond the simple fact that IoT deployments will place demands on traditional data center resources, whether it’s compute, storage or connectivity, those data centers are going to have to learn to handle all-new functionality. Where that’s going to hit hardest, according to Gartner vice president of analysis Mark Hung, is in management.
“Given the complexity of some of the new IoT applications, some of the operational and management functions also need to be revamped, using new automated/ML-based technologies instead of manual control,” he said.
That’s a big task, complicated by the fact that IoT deployments, for all their rapid growth, are still in the early stages, and nobody’s really sure how to characterize their overall impact on data-center operations.
“There is really no industry best practice or even rule of thumb that has been developed yet, as we are still in the early stages of IoT deployment,” noted Hung.