The near-term effects of the IoT/Cloud/ML convergence

January 26, 2018

By Cliff Federspiel, president and founder

The convergence of IoT, AI/machine learning and the cloud will impact commercial and industrial operations in ways we can’t yet imagine. But one thing that’s immediately clear is there will be way, way more data. The data itself will become its own force of change.

Fortunately, the availability of massive cloud storage makes it possible to collect and access this data, and making increasingly sophisticated data processing technologies like machine learning and analytics more viable – and necessary. As this data becomes networked and supported with smart software, more changes are in store. Here are some of the nearer term effects I foresee:

Data will become increasingly separated from its place of origin

With real-time data available from sensors monitoring virtually every aspect of system performance, facility operators can become physically separated from the assets they manage. When managers have access to holistic, real-time performance data and automated control, remotely controlled data centers become more viable. This, in turn, enables data centers to be located in areas which have cost or climate advantages, but which are less attractive to staff.

Conversely, data centers that must be located near urban areas for latency purposes, can have greater design and location flexibility. Microsoft demonstrated an example of this with its undersea data center pilot. DCD cited 10 “extreme” data centers, including locations in space and underground. Those locations will seem less extreme in the near future.

Alarming will become highly sophisticated and predictive

Legacy alarms are not informational. I’ve seen data centers in which a single common alarm rolls up 70 individual alarms. The consequence is that technicians get dispatched for non-emergencies and with little information on what’s happening, and what to look for.

With more detailed, granular information provided with alarm receipts enabled by the convergence of IoT, Cloud and machine learning, facility staff can readily discern whether there’s a need for immediate dispatch, or if they can simply add an investigation to an upcoming maintenance call. Predictive alarming will become standard. The quantity of actual alarms will decrease as the focus shifts from emergency response to alerts for corrective prevention – which is significantly less costly.

Reduction of alarm quantity, more predictive maintenance and complete information about the nature of the alarm will further advance the industry toward remotely managed data centers.

Maintenance will be become more predictive

Instantly accessible real-time and historical data on virtually every operation of every piece of equipment and its resulting effect on the white floor, makes declining performance easy to spot and address. IBM recently ran a Watson ad illustrating this very point. The ad shows an elevator repairman arriving at an office location to fix a problem which hadn’t yet occurred.

I’ve seen hundreds of data centers – some of them quite state-of-the-art, running with partially broken equipment and highly unbalanced airflow. These issues will become less common going forward. Facility managers can use equipment performance data to prioritize maintenance, reducing the frequency of actual equipment failures. Condition-based maintenance can become the norm. Maintenance managers will have data to verify if the work of their staff, or the work of outsourced third parties, was successful. Data centers will operate with less risk – even as operational complexity increases.

Security will change from air-gapped to built-in. The protocols and automation in place at data centers today aren’t secure. Air-gapping, the physical isolation between a secure network and unsecured networks, is the common security mechanism for data centers. This makes sense when data centers operate as standalone buildings. However, air-gapping and leveraging the benefits of a networked, cloud-friendly operation is incompatible.

Going forward, data will be networked. Legacy protocols with weak or non-existent security will no longer be sufficient. New IoT-based protocols and APIs already include modern security as a design standard.

Standard operating procedures actually become standardized

Corporate-wide, facility-to-facility conformance to standard practices for energy efficiency, maintenance and operations will become manageable thanks to data-enabled transparency and centralized operational control. Best practice policies can be enforced at higher management levels to ensure that financial and operational goals are met. Facility owners will be able to ensure that best practices are being followed, while reporting compliance to customers and executive management. Transparency increases control at all levels with the most dramatic effect on portfolio and management-level visibility.

A data becomes increasingly networked, data centers must prepare for this new normal. Data and the combination of IoT, cloud and machine learning convergence will drive and enable considerable operational optimization. Companies that move quickly to use this data to gain sights and take action will gain clear advantage.

There’s no question that this convergence is in play, only a matter of how ready we’ll be to leverage its benefits.


This article originally appeared in DataCenter Dynamics.