Last year I was part of a small team at Emerson Network Power that was asked to identify emerging data center archetypes based on what we were hearing and seeing in the industry. We looked at the disruptive forces that are driving change in the traditional data center and identified four emerging archetypes: cloud of many drops, the data fortress, the corporate social responsibility compliant data center, and fog computing.
While all address specific challenges data centers operators are facing, fog computing is likely to have the most widespread impact. In fact, if you work in manufacturing, mining, refining, processing or any of a number of other industries, fog computing may be unavoidable.
That’s because fog computing — a term coined by Cisco — addresses the challenges posed by the Internet of Things (IoT) — and, by almost all accounts, the IoT is going to be huge. Consider these projections:
- Gartner predicts that 25 billion connected devices will comprise the IoT by 2025
- IDC predicts the global IoT market will grow to $1.7 trillion in 2020 from $655.8 billion in 2014
- McKinsey estimates the potential economic impact of IoT technologies to be between $2.7 and $6.2 trillion per year by 2025
Those projections will only be realized if adopters of IoT can effectively manage the huge volumes of data being created and the current cloud-based model isn’t well suited to accomplish that. There isn’t enough bandwidth to handle all of the data being created and a remote, centralized data center can’t respond quickly enough to enable the productivity and downtime-reduction benefits that IoT adopters will expect.
Interestingly, this is similar to the challenge data center managers have faced in their efforts to centralize data center management. The volume of data created by servers, switches, power protection systems, and thermal management units is so large it can’t be effectively collected and transmitted in real-time directly to the management system in a way that enables rack-level visibility or system-level optimization.
The solution, as we’ve discovered in the data center, is intelligent controllers and gateways that collect data from devices operating in close proximity to each other or across devices that comprise a system. This local level of collection and control, which is what the fog computing model is built on, addresses several of the most significant challenges posed by IoT.
First, it provides a solution to the volume of data being generated by focusing consolidation on a location or system. It can also perform the processing-intensive work of translating data from different types of devices into a common protocol or language and analyzing that data. An intelligent control system operating at this level is also able to respond much faster to changes in operating conditions than a centralized system, enabling real-time machine-to-machine communication and system level optimization while providing a common human-machine interface. Finally, it filters data for the central repository where it can be analyzed using the computational horsepower of big-data systems in context with data from other IoT networks and archived.
When viewed in this context, fog computing is very similar in both principal and technology to a thermal management system that uses an intelligent controller to create a network of wireless sensors, room and rack-level cooling units, and an economization system to safely manage heat removal across a facility using the minimal amount of energy. The thermal management units and sensors comprise an IoT network within the data center in which units communicate through the controller to coordinate their response to changing conditions while the controller enables system-level access and monitoring and transmits relevant device and system data to the centralized management system.
Fog computing applies this approach on a larger scale to distributed IoT networks. Local intelligent controllers and gateways manage data consolidation and device control while also communicating essential data to the centralized data center. This will not totally isolate the data center from the impact of IoT — managing this network of networks will add complexity — but it does provide a viable solution to the significant challenges posed by IoT. And, employing technologies and specifications proven in the data center — which is a particularly demanding IoT environment — to remote networks can minimize the management challenges.