Data Center Modelling and Ice Storage

9 posts / 0 new
Last post

Hello all,

I am attempting to model a data center in eQuest for LEED. This will be a first attempt at modelling a data center for me and I am in hopes someone could provide some sage advice on the subject.

My understanding is that the data center will be a standard hot/cold aisle configuration with underfloor cooling (UFAD). The CRACs are on CHW and the CHW plant consists of multiple chillers, as well as, dedicated chillers for ice storage.

Recognizing that eQuest doesn't consider fluid air dynamics or temperature stratification resultant from UFAD, I am considering using the UFAD modelling guidance in the EDR Design Guidelines: HVAC Simulation Guidance (http://www.energydesignresources.com/resources/publications/design-guidelines/design-guidelines-hvac-simulation-guidelines.aspx)

This document addresses UFAD, but not specific to data center or hot/cold aisle. What Ive gotten out of the guidance is that for comfort UFAD applications, the upper portion of a room (non occupied ) space is defined as plenum and the internal loadings for the room are apportioned between the occupied zone and plenum. It seems to me that this method could be appropriate for hot/cold aisle scenarios as well, with the hot aisle as the plenum and cold as space. From the guidance equipments loads are suggested to be apportioned 67% space 33% plenum for comfort applications.

My understanding of hot/cold aisle configurations is that cold air is supplied on the front side of the server racks, pulled through the racks by the server internal cooling fans, and exhausted in to the hot aisle. Following from this logic it seems to me that the equip load should apportioned mostly to the plenum (hot aisle).

Does anyone have a better suggestion for load proportioning for a datacenter? Am I over complicating this?

My next challenge is modelling the Ice storage. I have 6 chillers of which 2 are dual mode CHW and Ice making. My guess is that I will have to model the dual mode units as 4 chillers (2 for CHW and 2 for ice) because of the different efficiencies and supply temps, while somehow using to the equip controls to make sure the chillers are not operating simultaneously.

This will be another first attempt for me. If anyone knows of any useful modelling guidance docs or has any advice on this it would be highly appreciated.

Thank you in advance and regards,

Shuichi Hendrickson, LEED AP

Shuichi Hendrickson's picture
Joined: 2011-09-30
Reputation: 0

Hi Shuichi,

Below is a link to an additional source from a previous post that helps
explain how to model UFAD and/or displacement ventilation. Their method is
very similar to yours, in that you split the equipment loads between the
plenum and the space. In particular, there is a section which explains
pretty clearly how to model UFAD in eQUEST. This may be repeat info to you,
but it may be able to help.

http://www.archenergy.com/ieq-k12/Public/Proj2_Deliverables/D2.9c_FinalDVDesignGuide_2006-0630.pdf

Best, Patrick

Patrick Keeney's picture
Offline
Joined: 2011-09-30
Reputation: 0

Side note that you may want to keep in mind that?s been brought up here is how you model the data center process loads. If the loads are too large, you may have a difficult time meeting the prerequisite.

William Mak, LEED AP BD+C

Will Mak's picture
Offline
Joined: 2011-09-30
Reputation: 200

Patrick, Will, John and all

Thank you all for the quick responses to my original post.

Based on your comments, I have a couple more questions that I hope someone may have insight into.

It seems that due to the large process equipment energy usage for datacenters, others have experienced that showing significant energy reduction through HVAC alone is difficult. Following from this, for the purposes of LEED it appears that is would be best to find some strategy to minimize process loads in the model or somehow show a process load reduction.

1) For the project I am working on, we are starting off with 3 X 600RT CWH chillers + 2 X 600RT dual mode (CHW and Ice) chillers on Day 1, but with an additional 1 x 600RT CHW + 2 X 600RT dual mode units are planned for full future design cooling capacity. Correspondingly, the Server (process equipment) will be installed in stages as well.

Does anyone know what if there are any rules as to what point in project development the model should be based on? My original plan was to model at full design capacity, but I am thinking now that it would be better to model Day 1 conditions if permissible to lower the denominator in the energy reduction calculation.

2) The client is considering using Server Virtualization technology for the IT equipment. I understand that this is a much more efficient means of operating server equipment and I am considering using this as a basis for showing process energy reduction based on a CIR I have come across (CIR 2441). The CIR was originally submitted for a remodelling project, so the process reduction would have been shown versus the originally installed equipment.

Does anyone have any thought or experience any with showing this type of process reduction for a LEED-NC project? I am imagining the baseline will be based on servers without virtualization. I don't really know enough about "Virtualization" at this stage to really quantify what energy reduction, if any, could be expected. Some long conversations with the IT designer are warranted, but would appreciate if any one has any thoughts to share.

Again, any thoughts will be greatly appreciated.

Thanks in advance,

Shuichi Hendrickson

Shuichi Hendrickson's picture
Joined: 2011-09-30
Reputation: 0

An HTML attachment was scrubbed...
URL:

Bruce Easterbrook's picture
Joined: 2011-09-30
Reputation: 0

Where does it say LEED does not deal with process loads? It states that you must model the process loads anticipated on the design on both proposed design and baseline models.

William Mak, LEED AP BD+C

Will Mak's picture
Offline
Joined: 2011-09-30
Reputation: 200

An HTML attachment was scrubbed...
URL:

Bruce Easterbrook's picture
Joined: 2011-09-30
Reputation: 0

Bruce and all,

Thanks for your input below, I will definitely take breaking out server space into additional zones per expansion phase into consideration.

Regarding process energy, my understanding is that there is a mechanism for showing process energy reduction via the exceptional calculation method per ASHRAE 90.1 ( See LEED-NC 2009 EAp2 pg 238 paragraph 2 and EAc1 p258 2nd to last paragraph of Option 1). Admittedly, I have never used an exceptional calculation to show process energy production, but it seems that it is possible. As mentioned in my second post, I came across a CIR that approved using Server Virtualization as means to show process reduction for a data center (Copied Below). We have submitted a CIR to confirm that this would be possible solution for the project, but have not received the response yet.

The CIR is for an existing data center so the baseline would have been existing equipment. In my case, I am operating under the premise that standard servers for the baseline would not adopt Virtualization. Though, to honest, I don't know if Virtualization for a new datacenter would actually save as much energy as claimed below. I think an argument could be made that Virtualization allows for higher performance machines that use more energy to being packed into the same amount of space. The energy reduction would be realized at part load when a single "Virtualized server" could support the tasks of several, allowing those units to be turned off. I don't know enough about servers to comment much beyond this at this time.

It sounds like you do a fair bit of datacenter modelling, so perhaps the above could be of value to your projects as well.

I would appreciate if anyone has any experience to share or comments on using exceptional calculations for process energy reductions.

CIR 2441
Ruling [ ]

The applicant may use the Exceptional Calculation Method to take credit for any energy savings available from the server virtualization technology. Be sure to include as supporting documentation under EAc1 all assumptions made in the calculations, detailed data, any actual measurements taken to support the savings claims and any other pertinent information. Please note that the actual amount of credit will be determined by the review team at the time of the review.

Inquiry [ ]

This project is a remodeling of an existing office building that will include the addition of a Data Center. The Data Center will make extensive use of server virtualization to save energy, space and money. Server virtualization is the technique of replacing multiple servers with one server running multiple "virtual servers" on one larger, high performance server. In their existing Data Center, our client runs an average of 8 virtual servers on each high-performance server. Although the high performance server uses more energy than any one of the low to medium performance servers that it replaces, this results in a net energy savings of between 60-75%. Our question is: Under EAc1, Option 1, may we use the Exceptional Calculation Method to model the energy savings accountable to server virtualization?

Thanks in advance and regards,
Shuichi

Shuichi Hendrickson's picture
Joined: 2011-09-30
Reputation: 0

An HTML attachment was scrubbed...
URL:

Bruce Easterbrook's picture
Joined: 2011-09-30
Reputation: 0