COVID-19: Critical business support and our focus on employee health. Learn More

The Raritan Blog

Avoiding Unscheduled Downtime by Removing the Guesswork from Rack Loading

Posted on June 18, 2014 by Jon Inaba

Traditionally, data center operators and IT professionals have used a very simple process to create a safe “A” side to “B” side failover scenario. The process and math are very straight forward. The user decides what a safe threshold is - typically 80% of the total kVA of the enclosure - and divides that value evenly between both rack power strips and their onboard circuit breakers.  Due to the relative simplicity of the architecture, this practice can work well in a single-phase, less than 5kVA environment.  However, in power kVA environments greater than 5kVA, and more complex three-phase environments that are now typical for blade and high density environments, this practice dramatically increases the likelihood of encountering a cascading failure and creates the additional downside of stranding large amounts of available space, power and cooling resources in the process.  

In order to be truly safe and avoid the negative side effects of stranding otherwise available resources, the data center operator must be able to precisely model their environment.  Ideally, this means having a data driven model that reflects both the max and average loading across all outlets, circuit breakers, and the inlet power lines.  Acquiring a usable level of precision and accuracy is extremely difficult for most operators for a variety of reasons, but the biggest challenge to overcome is the ability to correlate these data points across multiple power strips.

Even with the help of a clipboard and calculator, most of us simply can’t do the relatively complex mathematics necessary to understand the effects of additional loading or a failover in a three-phase system.  This is where employing intelligent rack PDUs and a power management system is highly advantageous. Rack PDUs can automatically collect real-time power information down to the outlet-level and provide aggregate power values across multiple strips. With this type and quality of data, commissioning is done with extreme accuracy and potential threats are eliminated. 

Watch this video to learn how some data centers use power management systems to find stranded capacity and improve data center efficiency.