Session Title Demystifying Efficiency in the Data Center

Session Title: Demystifying Efficiency in the Data Center Utilizing Airflow as a System Presented By: Jon de. Ridder Enabled Energy

Learning Objectives: • Identify how to improve your Power Usage Effectiveness immediately • Design an efficient airflow system in a data center and apply to your own facilities • Measure and verify the savings achieved in efficient data centers • Identify ASHRAE TC 9. 9 and its effect on the ecosystem of the data center

Background Defining the terms: • Data center – The room (regardless of size, age, how anyone “feels” about it, the budget that you have [or had] to build or maintain it, or how reliable it is / is not) that houses your computing equipment.

Background Defining the terms: • Data center • Computing equipment – Server, network, or storage devices that compute, transport, and store information (data).

Background Defining the terms: • Data center • Computing equipment • PUE – Power Usage Effectiveness. Taking the total facility power (feeding your data center) and dividing by your IT load (UPS load will get you close) will give your PUE. This PUE number will be greater than 1 (hopefully less than 3) and provides a “uniform” way of calculating how much power is going to your IT load vs. how much power you are consuming to accomplish your compute (the “tax”). A PUE of 2 is typical in a “legacy center”, while a PUE of 1. 5 is “typical” for a new data center build (many are now becoming very aggressive i. e. 1. 1 and 1. 2). PUE = 3. 0 IT Equipment 33% Air Movement 16% Electricity Transformer/ UPS 13% Lighting, etc. 4% Cooling 33% Courtesy of a very sad data center experience

Background Defining the terms: • Data center • Computing equipment • PUE • A tax – Something you pay because you are forced to or because you are not aware of it.

Background Defining the terms: • Data center • Computing equipment • PUE • A tax • ASHRAE TC 9. 9 – The American Society of Heating, Refrigeration, and Air Conditioning Engineers Technical Committee 9. 9 brought together many – – hardware manufacturers, locked them in a room, and came up with the latest “Thermal Guidelines for Data Centers”. This is a GUIDELINE NOT A STANDARD. Your equipment warrantee is provided by your equipment manufacturer and ultimately this is who gets to decide if you are or are not “compliant” with housing the equipment in a “proper environment”. Examples: power quality; temperature and humidity controls; particulate type; and size. In the end data wins…he or she with the most information is likely going to be the person who controls how, when, who, where, and why.

Background Defining the terms: • Data center • Computing equipment • PUE • A tax • ASHRAE TC 9. 9 • Reliability – The resulting investment of many painstaking strategy sessions (brain cells) coupled with lots of redundant components (which – translates to big dollars) allowing for the concurrent maintainability of your entire infrastructure (planned maintenance to avoid system downtime). Hope is not a strategy!

Background Defining the terms: • Data center • Computing equipment • PUE • A tax • ASHRAE TC 9. 9 • Reliability • Efficiency – An aggressive pursuit (and an exhausting effort after achieving the appropriate levels of redundancy) to achieve maximum – – – throughput with minimal restriction and waste. This starts with doing the best you can with what you have, but working intently and diligently to make it better. Please note that reliability is and must be first. Walnuts can be opened with steamrollers, but they don’t need to be and the result isn’t pretty.

Background Defining the terms: • Data center • Computing equipment • PUE • A tax • ASHRAE TC 9. 9 • Reliability • Efficiency • Problem – Opportunity Self Actualization Esteem Social Safety Physiological

Background Defining the terms: • Data center • Computing equipment • PUE • A tax • ASHRAE TC 9. 9 • Reliability • Efficiency • Problem – Opportunity Optimization Efficiency Reliability Strategy Communication

Airflow - A Systems Approach Cause: Effect: Meaningful metrics are needed for the data center. PUE and CUE are now metrics the industry is accepting as “standard” and yet these are not universally understood or defined. Cause: Effect: Delivery systems were developed around outdated guidelines. Dramatic overcooling of IT equipment! ASHRAE TC 9. 9 published new thermal guideline for data centers (~78. 6˚F at the intake of compute equipment). Cause: Airflow delivery systems are generally unbalanced and full of air-mixing opportunities. Typical delivery systems have >50% “bypass” airflow. Effect:

Airflow - A Systems Approach It all starts with § Organization § Distribution It falls apart with § Poor communication § Bad strategy

Discover Your PUE • Calculate how much you are spending now on the system and each part that creates the total. Your CRAC/CRAH efficiency • Start with the intake temperature of your server, network & storage equipment. • Then calculate the efficiency of your CRAC/CRAH units – (CFM * delta temperature) *. 9 = BTUs of accomplished cooling. The path for your airflow • Supply path – Supply panels – Aisle layout (hot/cold) – Opportunity for recirculation • Return path – What is the path of least resistance?

Most Valuable Investment (MVI)

Bypass Airflow Source: Up. Site

Most Valuable Investment (MVI)

Forward-curved blades use blades that curve in the direction of the fan wheel's rotation. It has 24 to 64 shallow blades with both the heel and tip curved forward. Air leaves the impeller at velocities greater than the impeller tip speed. Tip speed and primary energy transferred to the air is the result of high impeller velocities and operating most efficiently at lowest speed. Backward-curved blades use blades that curve against the direction of the fan wheel's rotation. The blades are single thickness with 9 to 16 blades inclined away from the direction of rotation. Air leaves the impeller at a velocity less than its tip speed. Relatively deep blades provide efficient expansion with the blade passages. The backward curvature mimics that of an airfoil cross section and provides good operating efficiency with relatively economical construction techniques. Backward-curved fans are much more energy efficient than forward curved fans. The EC Fan design moves the air in more of a straight line.

EC Fan Forward Curved Fan Reduced AIR FLOW AREAS 17”

Most Valuable Investment (MVI)

Under-Floor Baffle

Most Valuable Investment (MVI)

Proportional Distribution Tiles

Sealing Cable Cutouts

Most Valuable Investment (MVI)

Blanking Openings in Cabinets

Most Valuable Investment (MVI)

Containment

Most Valuable Investment (MVI)

Ducting CRACs to Drop-Ceiling Air Space

Most Valuable Investment (MVI)

Optimized

Thank You!

7 X 24 Fall Meeting Airflow secrets revealed Tom Weiss President C 2 October 23, 2012

Presentation goals § Provide baseline for analyzing effective airflow cooling in a data center § Provide financial measurements for airflow efficiency § Share common problems and their source § Provide information on effective CFD § Answer questions regarding airflow

Let’s start with the science Air cooling is a method of dissipating Heat! It works by making the object to be cooled have a larger surface area or an increased flow of air over its surface; or both!

Problems with some raised floor cooling! Advice from Eaton § “As much as 30 to 60 percent of the data center utility bill goes to support cooling systems. If that figure seems too high, it is. ” § “Poor airflow management reduces both the efficiency and capacity of computer room cooling equipment. ”

An easy fix? …. “optimize the existing cooling system through mechanical and room layout changes, using relatively inexpensive devices to redirect and concentrate available airflow”.

What can we fix? § Cold Aisle / Hot Aisle Containment § Curtains, blanking panels § Close the holes in the floor We still haven’t determined what happens to the air once it leaves the floor. So how does the cold aisle really work?

CFM vs. Usable CFM Traditional Measurements § CFM – Cubic Feet Per Minute § Static Pressure and open space Performance questions § How much air are we wasting? § How does the air flow out of the tile? Do we need more air or do we need to be more efficient with the air we have?

How should it work? The three components of cold aisle airflow efficiency: § Cool the upper servers § Flow to the servers § Come out of every section of the tile § No back flow into the floor!

Stratification heat!

Industry white paper §

Tile mixing!

Air is like water! § Bypass the servers § Entrainment § Pollute the return air § Room mixing

Stratification

Short cycle / Jet stream

950 CFM - Venturi

Wasted air = wasted energy + hot spots § “Only 28% of the air, in a traditional raised floor system, cools the servers (72% does not cool)” § “ 2. 6 times more cooling than is necessary and yet we still have hot spots” § Lower set points § More CRAC units § Higher energy cost Data from Uptime Institute Dr. Bob F. Sullivan and Kenneth G. Brill – “ 24 by Forever”

How expensive? Cooling costs are 50% of the total bill! “Data center managers can save 4 percent in energy costs for every degree of upward change in the set point, according to Mark Monroe, the Director of Sustainable Computing at Sun Microsystems (JAVA)”

Data Center costs § OPEX – Energy cost 40%! § Asset refresh costs – § Server failure 20% higher! § CAPEX– Crac units, more devices! § PUE – 2. 0 or higher!

What is going on? CEETHERM / NSF study § Measure relationship of airflow to rack inlet § Analyze air at the particle level § Analyze CFD models § Provide feedback to the industry

5 4 Tile/Rack Level Air Flow Modeling 05/04/2012

5 5 Rack and Tile Geometry Fan speed setting dial Perforated floor tiles with dampers Server Simulator

Pressur e Outlet 56 Details of Tile geometry Pressur e Inlet Tile top 30 30=900 pores, 0. 5”, Porosity=39% 8 8 Fan Grill Tile 4 million cells Symmetr y Ra ck Mass Flow Inlet Damper s 4 9=36 pores, 1. 25” 5. 25”, Porosity=41% 20 84

5 7 Details of Rack Geometry 12 inch 18 inch Fan Wall Server Simulator 1 Wall Server Simulator 2 Pressure outlet Side View 5 inch Wall 1 inch 1 ft Symmetry Front View (aisle) Front View (fan) Front View (grill) 0. 5 inch 3 inch 6 inch open area 1 inch wall Grill Wall Server Simulator 3 Wall Server Simulato r 4 Pressure outlet Pressure inlet hub Pressure inlet (K=10) wall fan Aisle 9 inch Tile top 1. 5 inch Dampers Plenum 24 inch Symmetry Mass flow inlet Tile top Dampers Plenum Symmetry 24 inc h Mass flow inlet 9. 25 inch 18 inch • Fan: Target mass flow rate boundary

5 8 Current CFD Models for Tile Flows • Porous jump model [1] Low Pressure Uniform Velocity P Tile (F) Uniform Velocity • Body force model [2] ü As mass is conserved hence momentum is under accounted: mass flow velocity in through porous jump model pores ü Momentum source term specified just above the tile: Vin High Pressure Velocity Profile S x X Uniform Velocity [1] Patankar, S. V. , Airflow and Cooling in a Data Center, Journal of Heat Transfer, 2010, Vol. 132, pp. 073001 -1 -17. [2] Abdelmaksoud, W. A. , Khalifa, H. E. , Dang T. Q. , Elhadidi, B. , Schmidt, R. R. , Iyengar, M. , Experimental and Computational Study of Perforated Floor Tile in Data Centers, Intersociety Conference on Thermal Phenomena (ITHERM), Jun 2 -5, 2010, Las

Rack Flow = 2594 CFM Tile Flow = 0 CFM (0% of Rack Flow) Tile Flow = 0% of Rack Flow Aisle top Velocity (m/s) Rack Aisle center Aisle top Tile PIV CFD [ref] Kumar, P. , Joshi, Y. , Experimental Investigations on the Effect of Perforated Tile Air Jet Velocity on Server Air Distribution in a High Density Data Center, Intersociety Conference on Thermal Phenomena (ITHERM), Jun 2 -5, 2010, Las Vegas, USA.

Tile Flow = 20% of Rack Flow Aisle top Velocity (m/s) Tile PIV Aisle center Rack Aisle center Aisle top Rack Flow = 2594 CFM Tile Flow = 496 CFM (~20% of Rack Flow) Tile CFD [ref] Kumar, P. , Joshi, Y. , Experimental Investigations on the Effect of Perforated Tile Air Jet Velocity on Server Air Distribution in a High Density Data Center, Intersociety Conference on Thermal Phenomena (ITHERM), Jun 2 -5, 2010, Las Vegas, USA.

Tile Flow = 60% of Rack Flow Aisle top Tile PIV Aisle center Rack Aisle center Aisle top Rack Flow = 2594 CFM Tile Flow = 1598 CFM (~60% of Rack Flow) Tile CFD [ref] Kumar, P. , Joshi, Y. , Experimental Investigations on the Effect of Perforated Tile Air Jet Velocity on Server Air Distribution in a High Density Data Center, Intersociety Conference on Thermal Phenomena (ITHERM), Jun 2 -5, 2010, Las Vegas, USA.

Tile Flow = 100% of Rack Flow Aisle top Rack Aisle center Aisle top Rack Flow = 2594 CFM Tile Flow = 2594 CFM (100% of Rack Flow) Tile PIV CFD [ref] Kumar, P. , Joshi, Y. , Experimental Investigations on the Effect of Perforated Tile Air Jet Velocity on Server Air Distribution in a High Density Data Center, Intersociety Conference on Thermal Phenomena (ITHERM), Jun 2 -5, 2010, Las Vegas, USA.

Tile Flow = 100% of Rack Flow Tile Aisle center Aisle top Rack Aisle center Tile Aisle top Rack Aisle top Aisle center Rack Flow = 2594 CFM Tile Flow = 2594 CFM (100% of Rack Flow) Tile Resolving Body Force Porous Tile Model Jump Geometry [ref] Kumar, P. , Joshi, Y. , Experimental Investigations on the Effect of Perforated Tile Air Jet Velocity on Server Air Distribution in Model PIV a High Density Data Center, Intersociety Conference on Thermal Phenomena (ITHERM), Jun 2 -5, 2010, Las Vegas, USA.

Let’s end with the science Air cooling is a method of dissipating Heat! § Object to be cooled have a larger surface area – § Freeze the room = $ § Or an increased flow of air over its surface – § Flow to the servers = higher set point savings! § Or both – § Flow of cool, non – polluted air, through the rack! Questions?

Thank you!

<Insert Picture Here> RMDC Cooling System Adjustments Kevin Donnelly Building Engineer

67

68

Example of Title Extending to Two Lines • First Level Bullet – Second level bullet • Third level bullet – Fourth level bullet • Fifth level bullet 69

Program Agenda Example • • • Our Understanding of XYZ Capabilities and Value Drivers Benefits & Assessments Oracle Solutions Oracle Credentials Appendix <Insert Picture Here> 70

<Insert Picture Here> Appendix 71

Temperature and Humidity Adjustment Results • • • Invested $31 K for installation of plenum return ductwork. Server inlet temperature reduced from 80 to 70 degrees Number of CRAC units humidifying from 11 to 1 Mechanical cooling energy reduced 155 KW (24%) Annual energy savings of $53 K Payback 7 months 72

73

74

Other Adjustments • Seal air gaps (Koldlok) • CRAC temperature setpoint 72 to 76 degrees • Condenser water loop setpoint 75 to 65 degrees 75

76

77

78

Questions? 79