DCIM – From Metrics to KPIs

American engineer, statistician, professor, author, lecturer and management consultant Dr. W. Edwards Deming once said

“In God we trust; all others bring data.”

While this is somewhat funny, the point is also very clear. Without proper data to back our ideas, strategies and journeys, we’re merely guessing and hoping for the best. However a very common challenge amongst data centre managers trying to mature the management of data centres, is best described similar to the Big Data challenge of other industries.

For years and years we have come acquainted to collecting all sorts of data from connected devices in the data centres. This could be power draw from a PDU or UPS, temperature and/or humidity from sensors (dedicated or “onboard”), state change from a generator, etc. With the explosion in the number of “smart” devices (a.k.a. monitoring-enable equipment) deployed in modern data centres, this means we’re faced with another challenge of making sense of all that valuable data.

My advice, when faced with this challenge, is always to find the common themes in the data. This can only be achieved by looking at the system rather than the individual data point. While data points are great for early warnings (i.e. threshold violations against a pre-defined ideal state), system level KPI data is better suitable as foundation for decision making. If you want to understand if the current “burn rate” of your data centre resources are on the right path, you need a whole range of data points summed up to provide the KPI you’re looking for.

And you need to drive this top-down, since the alternative will always fail. You simply cannot do a button-up based on your collected data and expect it to properly reflect your needs. Rather you need to start at the very top (or as close as you can get) by understanding what’s driving decisions in your organisation and start defining KPIs that supports that. Now cascade those down through the organisation and automate the feedback for each using your collected data points and display it using technologies like dashboards and/or scheduled reports.

Remember not everyone needs the same data or granularity of the data, so make sure you take the time to fully understand the needs of different parts of the organisation before announcing the KPIs. Also remember static KPIs are becoming increasingly rare (and for a good reason), so make sure the KPI setting exercise is a recurring effort.

As a second step make sure event data is proactively used to detect service issues, with a risk of negatively impacting your KPIs. A cable breath for instance could impact the uptime of your business critical service(s) and you need to know before it becomes an issue!

So, keep collecting all those data points, but also makes sure you proactively summarise them in KPIs that are relevant and important to your business.