Preventative maintenance is a term often used in the data center world, and has been the mainstay of keeping a data center running smoothly for the last 20 years. The reality, however, is that this method of maintaining reliability is being overtaken by the modern data center and its demands for reliability and permanent uptime. In fact, according to practical futurist Michael Rogers, data center uptime will be the top priority in the 2020s as a result of increasing pressures for available data and the proliferation of Internet of Things (IoT) devices.
The demands placed on the equipment in a modern data center are now much higher, as well. Although equipment reliability is improving every year, data center managers are facing increasing pressures to keep costs down while still maintaining constant uptime. As a result, equipment – such as battery strings and fire alarms – must be replaced before they fail in order to meet 100 percent availability. However, without real insight into performance, facility managers are likely replacing the expensive equipment earlier than unnecessary for fear of failure. In order to reduce costs, empower employees, and prevent unplanned failures, data center owners, and operators should consider using the data collected in their facility to proactively predict failure risks.
Generators in the Data Center
Data centers may be considered the “brain” of a company, but in the event of a power failure, generators are key to keeping that brain running and reliable. The generator is what holds the whole building together during a utility failure, and it is arguably the most critical piece of equipment in a mission critical facility. Despite their significance, some studies suggest generator failure plays a 45 to 65 percent role in unplanned data center power incidents.
In order to have visibility into potential failures, most facility operators use sophisticated monitoring systems that collect data via sensors in their electrical and mechanical infrastructure. In fact, the sensor business has exploded in the last 10 years with almost everything now attached to a sensor that can monitor, measure and quantify limits and levels in real time to provide important alerts and alarms. Typically, the alarm data gets reviewed, but what to actually do with all the data after it’s collected has not been of concern. While the data provides valuable information on what is currently happening, further analysis would enable facility managers to extract meaningful, actionable insights.
Prevention through Prediction
With the proliferation of sensors, data is already being collected from nearly every piece of equipment in a data center. Unfortunately, facility managers are doing themselves a disservice by not examining this data post-collection, as it holds information that could help data centers across the world ensure uptime by predicting equipment failures before they happen. In other words, preventative maintenance is no longer sufficient – especially as it relates to generators.
Having a generator serviced once a quarter and run every month may have been acceptable in the Y2K era, but now data center managers need to know the health of their systems before its too late to really ensure 100% uptime. To do this, they need to look ahead by using predictive analytics in order to flag potential issues before they occur.
Currently, facility managers look at their sensor information in real-time to check the current status of their generators. Are they getting too hot? Are they still running? However, the true value comes from storing the data and examining it. Analyzing the performance over time enables a facility manager to determine patterns that can be used to predict an outage before it happens. For example, tracking the fuel level and consumption rate allows for accurate run time predictions and the most effective refueling strategy. In addition, combining many of these generator parameters can help predictive analytics point to more subtle problems, such as determining energy conversion efficiency by examining kW output and fuel consumption rate.
Even better, sharing these findings with other data centers could provide the industry with a global view of the performance of every brand of generator. Facility managers can compare data, and assess hundreds of different kinds of generators, giving them specific and actionable insights into the equipment – as what might be too hot for one brand may still be appropriate for a different brand, or one dip in a certain brand’s fuel level may mean something entirely different to another.
Data centers have all this information at their fingertips, information that would significantly ease the mounting pressures they face for 100% uptime, yet facility managers don’t seem to even realize it.
While utilizing data collected in a facility is not currently an established business practice, I foresee within the next five years the science of prediction will be a mainstream staple for anyone operating mission critical equipment. The data center manager that gets ahead of the curve and begins using predictive analytics as a serious tool in his or her armory, is the data center manager that will remain in the light when others are fumbling in the dark.