Data Center Efficiency Optimization: Old and New Approaches

DCIM is part of the bigger data center re-engineering effort underway that ultimately deals with reducing the unit cost of WORK, while making it more predictable, defendable and supportable. (Hopefully every reader is thinking about that first sentence for an extra few seconds). In its current incarnation and likely for the next 5 years, DCIM is simply about using all of the knowledge available to apply discipline to the ongoing planning and operations associated with the data center. It is a perfect complement to the actual good physical design work that must occur for any data center to be successful. In the end, both good design choices and active management play a big part in efficiency, and efficiency directly affects the cost per unit of work.

There are two basic approaches to data center efficiency today; 1) Drive UP the utilizations for traditional IT, Power and Cooling components, and 2) Adopt new generations of equipment (facilities and IT) that have a much higher value over changing utilization conditions/demands.

Data Center Efficiency Approaches

Data Center Efficiency Approaches

In general traditional IT and facilities equipment runs increasingly more efficient as the demands upon them rise. For example, a traditional transformer-based UPS that is running at say 30% load is MUCH less efficient than the same UPS that is running at 60%.  How much different? At least 5% (85% versus 90%). So with existing gear, the goal would be to load these traditional UPS devices much higher. Allow them to run at 60% usage or greater and you’ll see tangible savings. That is simply a design change and one that any good facilities oriented engineer could undertake. The same thing holds true for the IT gear as well. Traditional x86-based servers take 60% of their rated current just by powering-ON, and then incrementally consume more as the load increases. Again increasing the utilization of the gear is critically important, so VMware, Microsoft, Citrix and other have all devised very capable approaches to virtualization to increase the utilization of these servers from 10% to 70%. In this scenario, you simply get more work per watt, using the same type of equipment, just used in a much smarter way.

BUT, there is a new wave not just coming, but in most cases already here! Old equipment designs are being replaced by fundamentally new designs, both for Facilities and for IT gear. Consider the UPS mentioned above. For many applications, there are new solid-state UPS devices that run at greater than 95% efficiency, starting at just 20% utilization. This almost guarantees that the overhead associated with UPS losses is tiny. Similar advances have been made across many of the power and cooling components.

And for IT gear? You bet. ARM processors for volume webpage servers, variable speed fans in dense/blade chassis, high efficiency power supplies, low-power network optics, etc. Even Intel and AMD are talking about upcoming server designs that will decrease the base power requirements for a server from it’s current bare metal draw of 60% down to a more reasonable 15%-20%! What a perfect complement to virtualization for servers, storage and networking.

I see all of these approaches growing in momentum, and as such there isn’t one perfect answer that can be applied universally to each situation. But think of the possibilities when you combine new technologies with new designs, and active management. That said, this is a big grey area and underscores the need to work with highly trained data center and IT designers that stay up on all of the latest technologies. Think of it like a dentist that was trained in 1980, versus the one that just graduated. You really want a data center designer that is staying very current, and not simply replicating the designs that they did in 2005. As a general rule, ANY data center practice that was appropriate in 2005, is likely not nearly as desirable in 2013! And don’t forget DCIM Suite software which ultimately helps manage all that physical change. Change due to technology, and change due to financing and value. Combined with new technology, change must be actively managed and DCIM suites do just that.

Remember, we all have a lot of wrenches, it’s just knowing how to swing them that makes us geniuses!

About admin

Mark Harris Fremont, CA
This entry was posted in Uncategorized. Bookmark the permalink.

Leave a Reply