Blog

Note: This is part two of a three part series.  If you have not yet read part one, please click here to read do so before continuing.

Now that you understand the background behind the Rack Cooling Index (RCI) metric and why it is important, we will look at the Return Temperature Index (RTI) metric in detail.  The purpose of RTI is to assist data center operators in understanding the efficiency of the air handling system for the entire data center. The RTI metric is fairly sophisticated and needs much more information to compute than the RCI metrics.  However, it does provide valuable insight as to how the air handling system is performing.  Before we get into the specifics on RTI, lets examine and define two core concepts in data center air handling – by-pass air and recirculation air. 

First a little background on basic data center air handling.  Modern data center IT equipment is designed to pull in cool air from the front, use it to cool itself, then eject hot air out the rear.   Therefore the most common design for data centers is to cool air via CRACs then move the cool air to the front of the racks (air intake) and then evacuate the hot air (air exhaust) from the rear of the rack.   The exhaust air is then moved back the CRACs where it is cooled again and the process starts over.   

part 2 figure 1 graphic resized 600By-pass air is defined as air that does not participate in cooling the IT equipment in the rack.  The air somehow finds it way to the exhaust side of the rack without having to pass through the rack.  

part 2 figure 2 graphic resized 600The most common causes for by-pass air are:

  • Too much supply airflow (high air pressure) from the CRACs

  • Misplaced perforated tiles (in the hot aisles)

  • Leaky cable penetration cutouts in the floor tiles

The problem with by-pass air is that it is wasting cooling resources.  A good analogy is running the air conditioner for your car but leaving all of the windows and doors open at the same time. By-pass air equals wasted money. 

Recirculation air is defined as air that participates in the cooling of the IT equipment multiple times without being re-cooled.  In other words, the hot air from the rear (exhaust) of the racks is finding its way back to the front (intake) of the rack and is being pulled back through the IT equipment.  

part 2 figure 3 graphic resized 600

The most common causes for recirculation air are:

  • Too little supply airflow (low air pressure) from the CRACs

  • Lack or misplaced perforated floor tiles in the front of the racks

  • Lack of blanking panels on the front of the racks

  • Gaps between racks

  • Uneven rack row lengths (short rows and long rows intermixed)

The problem with recirculation air is that it warms up the temperature of the cool air that is intended to cool the IT equipment in the rack.  A good analogy is trying to cool your car by turning on the vent allowing outside air in to the car but not turning on the air conditioning unit itself.  The result is simply recirculating the hot air.   Recirculation air puts the IT equipment at risk of heat related failures. 

The RTI metric is a measure of net by-pass or net recirculation air. It is the ratio of total equipment airflow to total air-handler airflow expressed as a percentage.

  • 100% means balanced airflow

  • <100% indicates net airflow by-pass

    • Cold supply air returns directly to CRACs without cooling the IT equipment:  Elevated fan power, reduced chiller efficiency

  • >100% indicates net airflow re-circulation

    • Hot IT equipment exhaust is drawn back into intakes: Hotspots and reliability concerns.

    • Often results in low RCIHI

part 2 figure 4 graphic resized 600Ideally, an optimized data center would have an RTI metric value of 100%.  However this is unrealistic.  In the real world, some amount of by-pass air and recirculation air will always exist. The point is to minimize these conditions therefore increasing the efficiency of the data center air handling system. Just like with the RCI metric, the RTI metric is a simple to understand metric that helps to explain the balance of the air handling system.   If a data center is outside of the target parameters of the RTI metric (<80% or >120%), then energy dollars are being wasted. 

Note: There is only one RTI value for the entire data center since it is a measurement of the air handling system as a whole. 

As I stated earlier, the RTI metric requires much more information to calculate than the RCI metric.   In order to calculate RTI, the following is required:

  • Rack level instrumentation (sensors)

    • Optimal: 3 temperature sensors (top, mid, bottom) front & rear of every rack

    • Minimum: 3 temperature sensors (top, mid, bottom) front & rear of every 3rd rack

  • CRAC level instrumentation (sensors)

    • One temperature sensor on the CRAC supply and one (or more) on the CRAC return

    • Each CRAC unit must be modeled and the typical CRAC volumetric airflow rate (static value) for the CRAC must be known

Now envision a data center that has been properly instrumented so that both the RCI metric and RTI metric can be properly and accurately calculated.  The immediate value of the system is to provide a real-time baseline value for each metric.  This is the beginning grade or scorecard. These metrics provide a powerful and informative view enabling data center operators to make changes and clearly understand the results of the changes.  At this point, the data center operators can go into a continual improvement and evaluation mode that is how true optimization is accomplished.  Furthermore as changes are made to the data center that are not optimization related such as equipment upgrades and higher compute loads, the results of these changes can be analyzed in real-time as well. 

Now think back to Part 1 of this blog series – specifically about the tangible benefit that for every 1 degree F that rack air intake temperatures are raised, 2% of the annual power costs can be potentially saved.  With the proper instrumentation, software, RCI metrics, and RTI metrics, these savings are attainable without introducing the risk of heat related equipment failure.   RF Code’s wire-free sensors and enterprise monitoring software enables this type of energy savings and optimizations.  Don’t try to optimize your data center in the blind!  

Rack Cooling Index (RCI) is a Registered Trademark and Return Temperature Index (RTI) is a Trademark of ANCIS Incorporated.  (www.ancis.us).  All rights reserved.  Used under authorization.