A lot of people are fretting over the Internet of Things and how it will affect enterprise infrastructure, particularly networks.
The specter of millions of data points sending continual streams back to the enterprise for capture, cleaning, analysis and downstream processing is enough to make any CIO cringe. It is also one of the reasons why so many organizations are looking to shore up their scalability and flexibility capabilities through virtualization and software defined networking (SDN).
The good news is that this notion of an information deluge hitting the data center is largely a myth, provided you get your architecture in place first.
As Ganesh Moorthy, a thought leader at analytics firm Mu Sigma, noted to DataQuest recently, the IoT’s impact on centralized data infrastructure should be minimal. In fact, he says you can expect volumes to actually decrease in the data center, although increase across the distributed enterprise. The reason is simple. The IoT requires fast turn-around of both data and analytics in order to capitalize on the many ephemeral opportunities the streaming data represents. This could be the profiles of shoppers in a store or on a website at any one time, or the traffic patterns of a metropolitan roadway during rush hour. Regardless of the size and nature of the network, results will come too slowly if all the data has to be shunted to a centralized location, processed, and then kicked back to the field where it can be of use. This is why most IoT architectures rely on increased compute capabilities on the edge, thus preventing centralized resources from becoming overwhelmed even as overall traffic is on the rise.
Centralized resources will still play a role in the IoT, says Ovum’s Gary Barnett, but it won’t be doing the heavy lifting in the form of massive bulk processing. Indeed, with 20 billion connected devices expected by 2020, even Amazon would have trouble with the load if it had to pull it all into one giant analytics engine. But if you aggregate the feedback from even 100 sensors into a single, reportable average, you start to approach a data load that can be reasonably accommodated by a centralized, although highly scaled, architecture. Even then, 90 percent of the sensor-driven data will likely remain on the edge. At the same time, distributed resources will have high levels of intelligence to provide self-healing, self-tuning and automatic recovery and failover, reducing upstream traffic and monitoring as well.
The upshot of all this is that with the proper stratification of data infrastructure, network loads should not become overly burdensome, although network flexibility will be a premium asset, says Compass Datacenters CEO Chris Crosby. Speaking at the recent Data Center World conference in Las Vegas, Crosby outlined an ecosystem that can handle not only the burdens of the IoT, but increased mobile traffic, rich media streaming and a growing service-based economy as well. Naturally, this isn’t going to work with the standard Band Aid® approach to network development. It instead calls for a fundamental reimagining of networks and network architectures, with heavy reliance on NFV, SDN and VPN technology. Again, the key is flexibility, not bandwidth or throughput, and success will be determined by how well networks can reinvent themselves to accommodate shifting data loads, not how much raw data they can support.
This is also why you are going to see the term “fog computing” more frequently as Big Data and IoT architectures take root, says Data Informed’s Jelani Harper. Through decentralization and increased capabilities at the edge, the enterprise not only reduces the burden on key pieces of infrastructure but actually improves responsiveness and data productivity while reducing costs in the bargain. To be sure, there are challenges to this model as well, such as the need to drive high levels of intelligence and autonomy to the edge in the first place, not to mention harnessing it all under a common security architecture, but the benefits of faster reaction times and an improved user experience should be well worth the effort.
But this is no time for the enterprise to be sitting on its hands. The IoT is already operating in limited production environments, and with the potential benefits so enticing, it won’t be long before it becomes a common facet of the digital economy.
And it goes without saying that laying out network infrastructure before the IoT becomes a reality will be much easier than reconfiguring a patchwork of solutions on the back end.