If you’re running a data center, there are certain environmental and energy thresholds you need to stay in line with. If you don’t have the proper operating environment, that not only drives your operating costs through the roof; it could very well cause irreparable damage to your servers. Further, if you’re using too much energy, that’s a sure sign that your data center’s suffering from some glaring inefficiency or operations issue- which, ultimately, will hurt your bottom line.
If you’re a smart operator, these aren’t really things you can ignore.
So what environmental thresholds should you should for? What’s the ‘sweet spot’ for power consumption? How are they tied together?
The answer’s not actually as simple as you might think.
While it’s true that every data center should be both cool and reasonably dry, there’s a wide range of additional factors that come into play when you’re trying to determine the ideal operating conditions for your center:
- How old is your equipment? Older equipment is more sensitive, and tends to run hotter.
- What sort of server density is your data center running? As a general rule, more density means more heat.
- How many hours per year do you run your servers? Believe it or not, not every Data center in the world is “always on.”
- How intensive are the tasks you assign to your servers? Computers generate more heat when handling more intense tasks, after all.
Given all the above factors, a lot of operators err on the side of caution and run their centers at around 60 degrees Fahrenheit. Believe it or not, they actually don’t need to run that cold- ever. (unless you’re still using legacy hardware) Many operators are warming things up, raising temperatures to somewhere between 65 to 75 degrees. There’s the sweet spot, and if your data center’s running in that temperature range, you’re very likely in the clear. If it climbs above 85 degrees, you may have a problem.
Don’t err on the side of caution here, either- running too far below the threshold can actually end up costing you a pretty penny in climate control, and there’s a good chance that’s money you don’t really need to waste.
As far as humidity’s concerned, too dry is almost as bad as too wet. Ideally, you’re going to want to shoot for somewhere between 45 to 55%. Too damp, and it’s pretty obvious what’ll happen- computers and water don’t really know how to play nice with each other. Too dry, and there’s a good chance your gear is going to fry itself as a result of electrostatic discharge.
Power consumption’s a little trickier, but at the end of the day, it’s tied directly to efficiency. Elements which affect power consumption include:
- Climate Control- choose an inefficient or less-than-ideal means of controlling the environment in your data center, and you’re pretty much guaranteed to have your energy consumption shoot through the roof- and well above the threshold.
- The efficiency of your hardware- less efficient hardware will, naturally, use more power to accomplish less.
- The uptime of your facility- this one goes without saying. The longer your uptime, the more power you’ll be using.
As far as power thresholds are concerned, we’re going to swing back to my previous article- we’re going to look at the metrics behind power consumption. This is, quite simply, the only way we can really establish an energy threshold. One metric, in particular, is of interest to us- power use effectiveness. The closer you are to 1, the more efficient your center is- and the closer you are to the ideal threshold for power consumption.
You should also keep an eye on energy provisioning. Don’t waste power- you want to be absolutely certain that you aren’t provisioning too much power, or too little. Either way, it’s likely going to cost you money. Make sure you know how much power your center uses, and ensure you aren’t provisioning for more.
Environmental and Energy thresholds are both incredibly important factors in the smooth operation of any data center. Run above either threshold, and you not only waste money- you risk causing potentially irreparable damage to your hardware, leading to downtime that could end up costing you much more than a few million dollars. Thankfully, if you plan ahead, they’re both pretty easy to keep track of- just implement some means of monitoring them, and work out some threshold rules.
Questions? Comments? Concerns? Swing by our forum to have a chat.