Supernova Award Category

Data to Decisions

The Problem

Vertiv customers in their efforts to improve uptime and reduce risk have relied on direct services for their infrastructure support systems such as Uninterruptable Power Supplies (UPS).  These UPS systems provide backup energy for data centers when inbound AC power is lost.  Reliable power and cooling are critical to maintaining continuous availability of the world’s digital communications and economy.  

Until recently, field data provided by our customers was stranded in dozens of purpose-built systems.  For example, field electronic forms were in an island of information pertaining to startup and repair, while IoT systems held telemetry and alarms.   More than a dozen independent systems, both legacy and contemporary, performed their function and purpose well but did not interoperate or share data from one system to our staff and customers.

The Solution

Vertiv chose to utilize a high-performance Hadoop based data lake to combine all customer provided field data into one system.  We stream slow moving data such as software revision versions or model numbers exactly the same as fast moving data such as telemetry and alarms.  The combined model permits a blending of information so that each system contributes maximally for our customers, creating a “digital twin” of the system in the field.  For example, our electronic form system is the only system that contains an infrared image of temperatures, while our remote telemetry system’s data is unique at it provides internal temperatures.  Together,  these digital twins derived from blended data sources provide higher value than the sum of the individual readings.

The results

The results of our efforts have enabled new insights into how our customer’s products operate in the field from their first startup through their entire lifecycle.  For example, each of our UPS systems has a number of batteries, all connected in series like Christmas lights.  A single battery failure results in power failure to their data center or critical facility. The combination of electronic form reporting plus customer telemetry data have permitted Vertiv to be the first in the world to forecast battery service life for UPS batteries.   Additionally, from this we have incorporated health scores so that we can help our customers identify the most important systems to service first.

Metrics

In two short years the Vertiv data lake went from zero to 53 billion rows of data, with battery data from 250,000 in person inspections, plus 25,000 UPS and computer room air conditioning (CRAC) service interventions.  This has enabled ourselves and our customers to become more effective and reliable.  To date, not one customer utilizing blended field data and electronic reporting preventive maintenance battery inspections has lost power to their data center load.  Perfect performance.

The Technology

We utilize Cloudera Hadoop for our big data lake, Megger’s PowerDB for field inspections and Vertiv Life™ and Vertiv Alber for remote telemetry.

Disruptive Factor

The internal teams that support customers have dramatically altered their interactions with our customers after near real-time status of systems and field interventions (startup, preventive maintenance or repair).  Before reaching out to customers in any way, our inside teams can see all reports for all time and up to the minute status for operational performance.  Today this information is now being used to predict service life and guide customers in areas where unrealized efficiency gains are possible, saving energy.

Shining Moment

Most recently a customer asked us to prove that our data science, machine learning and methods performed as advertised.  We examined that customers entire portfolio and subsequently identified systems that were at risk.   The identified systems were considered healthy with classical monitoring technology.   What our customer thought was a healthy system with 15-20 minutes of runtime failed within two minutes of starting the test.