Prediction Is Better Than Cure – CFD Simulation For Data Center Operation

logo_ashrae

Prediction Is Better Than Cure – CFD Simulation For Data Center Operation.

This paper was written to support/reflect a seminar presented at ASHRAE Winter meeting 2014, January 21st by Mark Seymour, Future Facilities

Introduction

Advancing technology has brought us more powerful computer hardware for our data centers. In addition, it brings the opportunity to instrument and monitor the Datacom halls to better understand where the IT load is and the resulting environment, at least in terms of air temperature at an array of selected locations.

Instrumentation and monitoring is, without doubt, a step forward that helps the operator in their quest to understand and control energy use of their undoubtedly complicated asset. But is measurement enough?

Data center operators typically populate their white space slowly over time. They do so
assuming that the design intent (characterized by very high level parameters such as Total IT kW and IT kW/cabinet or IT kW/sqft) is always valid.

The problem with this assumption is that it does not reflect the reality of the operational configurations that will exist in the future. In essence, a data center designer has a challenging task – to design:

  •  The infrastructure for a room of unknown and varying electronics
  • With unknown and varying load
  • To be placed in an ill-defined climate – the local climate varying throughout the year.

So, the designer can only consider:

  • A range of design scenarios for simplified configurations e.g. Day 1 50%, 100%, using generic assumed loads and selected ambient conditions
  • A range of failure scenarios to check the design is resilient in the event of the unexpected

One thing we know is this: the configurations considered by the designer are unlikely ever to occur in practice!

This is not to say that CFD used in design to assess the design in relation based on these
assumptions is futile. On the contrary, such an assessment is the key to checking, selecting and optimizing the underlying concepts and strategies.

Further, it allows sensitivity studies for variations in IT load density, IT equipment type and configuration… in order to avoid design flaws that are subsequently exposed by small
deviations from the design assumptions during operation. However, such design assessment can never be considered a true prediction, because the configurations that will occur over time will be unique to that facility and probably to a particular day.

Given that these conceptual design simulations do not and cannot guarantee performance in normal operation, are the advances in measurement and monitoring the saving grace?

Download PDF here: ASHRAE_Prediction_Is_Better_Than_Cure

Table of Contents
Introduction
…………………………..
…………………………..
…………………………..
…………………………..
………..
1
Why Do Data Centers Fail to Achieve The D
esign Goals?
…………………………..
………………………..
3
Figure 1. Room Filled with Notional Equipment to Reflect Design Assumptions
………………………
3
Figure 2.
Uniform Load: All Cabinets 4.25kW / Cabinet (408kW total)
…………………………..
……….
4
Figure 3. Uniform Airflow Requireme
nt 240l/s/cabinet (23m
3
/s total)
…………………………..
………
4
Figure 4. All IT Equipment Operates In the ASHRAE Temperature Compliance Recommended
Range
…………………………..
…………………………..
…………………………..
…………………………..
…………..
4
Figure 5. Typical Enterprise Varied Equipment Configuration
…………………………..
……………………
5
Figure 6. Cabinet Loads Vary from 0kW to 13.2kW
…………………………..
…………………………..
……..
5
Figur
e 7. ASHRAE Temperature Compliance Is Not Achieved For Varied Equipment
………………..
6
Do Core DCIM Tools Address This Loss In Performance?
…………………………..
……………………….
7
Prediction Is Better Than Cure
…………………………..
…………………………..
…………………………..
………..
8
Simplifications In The Modeling Toolset
…………………………..
…………………………..
………………….
8
Figure 8. Sma
ll Datacom Hall with 8x No. 5kW Cabinets
…………………………..
…………………………..
8
Figure 9.
Comparison
of Elevation of Temperature In Cold Aisle
…………………………..
………..
9
Figure 10. Flow in the Raised Floor Predicted by RANS CFD (Left) and PFM CFD (Right)
9
Figure 11. Flow
from
In
Row Coolers Predicted by RANS CFD (Left) and PFM CFD
(Right)
…………………………..
…………………………..
…………………………..
…………………………..
……….
10
Simplifications/Assumptions When Creating the Model
…………………………..
…………………….
10
F
igure 12.
Temperatures
When 1U Server Are Above Blade Center
…………………………..
…..
12
Figure 13.
Temperatures
When 1U Server Are Below Blade Center
…………………………..
…..
12
Modeling for Operation
…………………………..
…………………………..
…………………………..
…………………
13
In Conclusion
…………………………..
…………………………..
…………………………..
…………………………..
……
1
4

Download PDF here: ASHRAE_Prediction_Is_Better_Than_Cure

About DCIMdatacenter

http://www.linkedin.com/in/rfschmidt
This entry was posted in Data Center DCIM Datacenter Datacenters Datacentre and tagged , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , . Bookmark the permalink.

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 )

Twitter picture

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

Facebook photo

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

Google+ photo

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

Connecting to %s