House loads / PUE

Right now we’re having a bit of a debate on how to calculate PUE. We are all familiar with the GreenGrid definitions of PUE 1, 2, 3 etc but our debate is which is most valid. As a colocation provider we lease data center space. We also lease data center adjacent office space. Should this space count against our PUE? Some of our sites don’t have offices while some do. Comparing them to identify opportunities would require a similar metric wouldn’t it? The office really isn’t supporting the center, it’s its own entity. At the same time, we have been including office, so the historical readings will no longer be as relevant if we change PUE. I’d like to open up a discussion here if anyone is interested.

I’m not as obsessed with conforming to an industry standard as I am with providing useful standards for comparison and progress. Of course us in the know are looking at so much more than just PUE, we break down the load path, find out where our power is and where it should be, and look for opportunities or solve issues. Yet PUE is an important tool for executive level reporting.

18 thoughts on “House loads / PUE”

    1. Sure it does. But can’t we measure that the way we measure the other 50 million square feet of office space in the CTL portfolio? What’s the PUE of an office? Often times, that office space has its own HVAC equipment as well. So now the equation is ((Cooling of IT space) + (House loads) + (Cooling of office space) + (Office space loads)) / IT Load. That numerator is pretty busy. And when we conduct a project to improve IT efficiency, won’t the change in PUE be underselling the results to a degree because of all the other things bogging down PUE? Then the executive level might see it and say “We installed EC Fans in Santa Clara and went from 1.60 to 1.45 PUE, but in Denver we did it and only went from 1.65 to 1.58, what did you guys do wrong?”.

      Like

    2. For internal benchmarking we can always define our own PUE, what to include or exclude, where to measure etc. But once it gets published to outsider, people will start to compare and challange how you arrive at that. A difference between 2 and 1.5 is obvious, but a lot of people will compare 1.5 with 1.55, thye don’t understand PUE and don’t realize they can’t make that comparison due to the inherit difference in measurement in different data centers.

      Like

  1. As you state the office is its own entity and being such my opinion would be that it should not be included in any part of the PUE calculation. A gray area that may arise is when the cooling infrastructure for the data center floor is what also supplies the office space. Office areas with their own cooling equipment are easy to eliminate from the PUE and should be. Offices that are supplied from the data center infrastructure will require more time and effort, but in order to get the most accurate PUE should also be factored out. After all PUE is for how efficiently a data center uses energy and I would say the envelope for that can be sealed, so to speak, at the raised floor.

    Like

  2. As Travis stated, if the office space uses the same cooling infrastructure then I believe it should be included into the site PUE.

    Another point of view could be, assume offices are “mini-datacenters” with personal laptops, screens, etc. as the it-load. Convert to this thinking portfolio wide and now you have just converted 50mil SQFT into trackable/power optimizational space, if it wasn’t already. As a datacenter company, that looks intriguing.

    Like

      1. The way is sub metering and calculating a cooling uplift based on known cooling loads. Another thing, we aren’t always cooling. We can use waste heat to heat offices in winters, but we (CTS) typically do not. Instead, office rtus have electric, inefficient heaters. If our waste heat went to the office instead, wouldn’t it make sense to have a positive variance for that benefit?

        Like

      2. According to TGG definition waste heat recovery should not go into PUE calculation but is expressed in another metric ERE (Energy Reuse Effectiveness).

        Like

      3. If we accept tgg as how we quantify PUE but we don’t really. We don’t specify if it’s PUE 123 or 4, we don’t mention if it’s average or instantaneous. Most people aren’t familiar with ERE either. It’s really important to quantify all the benefits of our actions. And you can’t help but not include waste heat use if the office load is where the heat goes and the kW not consumed in heating because of it would have went in the PUE numerator.

        Like

      4. For internal benchmarking we can always define our own PUE, what to include or exclude, where to measure etc. But once it gets published to outsider, people will start to compare and challange how you arrive at that. A difference between 2 and 1.5 is obvious, but a lot of people will compare 1.5 with 1.55, thye don’t understand PUE and don’t realize they can’t make that comparison due to the inherit difference in measurement in different data centers.

        Like

  3. Thinking in that way also opens the idea of, could the datacenter run without the office space there, or the employees that are housed in that space?

    Like

    1. Well technically, depending on which PUE you use (1,2,3 or 4) you have to evaluate if office loads are considered in the PUE calculation. More to the point, the offices used by facilities and operations teams exclusively supporting the data center infrastructure should be included with that PUE. So now we have another gray area of this office vs that office.

      Like

  4. I think if a company sets a PUE reduction percentage goal across the portfolio it could hold up as a good benchmark. But for example, to say every datacenter needs to be below 1.35 is a bit unrealistic. Because as we all know, a lot goes into factoring the PUE (location, climate, facilities, etc.)

    But if the main goal is to reduce energy consumption. Then the centers should start looking at improvements in IT energy efficiency not just the facilities side.

    Like

  5. PUE is used for benchmarking the facilities efficency of a data center over time, not to compare between different data centers as you can never make a fair comparison. Including office space would capture a more complete picture that encompass any good practice that enhance office space energy efficiency, but whether it is included or not included it should be done consistently for each data center for correct benchmarking.

    Like

    1. While we can’t compare data centers effectively because of the many differences, PUE lets us come close. It’s generally understood the a PUE of higher than 2 isn’t great and less than 1.5 is pretty solid. So in comparison with other centers, which we just can’t help but do, don’t we want to give each center the best story to tell. The centers will be compared by the people with the least understanding. Why hold back a building because we sold a ton of office space?

      For benchmarking over time, I prefer to look at the loads, cooling kW per IT kW, ups losses per IT kW, crah kW per IT kW, etc. those trends can provide valuable insight to reduce demand and identify the outliers which may be impacting efficiency.

      Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s