Energy Benchmarking Service for Data Centre Industry in Hong Kong and Initial Findings Simon Lam Acting Deputy Director, Business Marketing & Development Department, CLP Power
Agenda 1. Introduction: Why Energy Benchmarking? 2. Methodology 3. Initial Findings 4. Best Practices from the Top Performers 5. Way Forward 2
Introduction: Why Energy Benchmarking? Customer needs - DC is Energy Intensive - Keen to save energy cost - Premium customer service CLP s commitment in pursuing green - Responsible energy provider - Driving energy efficiency - Benchmarking tools in place - Eco-Optimizer (Domestic) - Energy Billboard (Retail/SMEs) World trend in energy saving - Benchmark with peers - Facilitate Behavior Change - Interactive product/service CLP s unique position - Independent position facilitates data & insights collection - Follow up with individual DC by dedicated account managers 3
Methodology Data Collection by Questionnaire Respondents were asked to submit data relevant to their energy consumption and, Describe their energy management initiatives Data Analysis, Verification and Insight Collection Telephone calls and meetings were conducted to verify reported data Following initial analytics, top performers were interviewed to understand their best practices (mainly facility load) Energy Benchmarking Report for Individual DC Customer CLP Account Managers visit individual customer and present the findings CLP can also offer energy audit services DC energy benchmarking is a complex exercise as there are many factors affecting DC energy efficiency. The service will be developed in stages, we start with quick win 4 items (e.g. PUE comparison) to build the momentum with customers.
Road Map of New Energy Benchmarking Service Stage ONE Stage TWO Value to Customers: Understand their energy performance amongst peers Learn the best practices Individual report Face to face discussion Invite more Data Centres to join Measurement Base lining Development of Energy Benchmarking service Concept development Data gathering Analytics & Validation Meeting with Top performers Present initial findings Result discussion Sharing of best practices in industry Collect more data and enhance the service Jan 2013 June September 2014 We are here today 5
Initial Findings 6
Overview Response rate: 30% Industry Type Client Type Institution, 10% Government, 10% Both, 13% IT, 6% Public Services, 19% Finance / Bank, 16% Co-location, 39% End-use, 48% DC Colo, 39% 7
Overview Tier Level Electrical System Generator Back-up Tier-3, 55% Tier-4, 6% Tier-2, 39% 2N, 67% N+1, 33% N, 0% 2N, 21% N+1, 28% N, 52% Use of Floor Area Chiller Type Non-IT area (facility), 38% IT area in use, 50% Both, 14% Watercooled, 28% IT Expansion area, 13% Air-cooled, 59% 8
Reported Power Usage Effectiveness (PUE) DC Industry standard to measure how well DC s manage their energy efficiency PUE = IT consumption + Facility consumption IT consumption (the smaller the better) PUE 4.0 3.0 Average PUE = 2.2 DC Avg PUE Age < 5 years Watercooled chillers IT Load kw/rack 2.0 Top 25% 1.7 50% 70% 3.2 1.0 Rest 75% 2.4 30% 30% 1.4 0.0 DC Sites Delve into 4 different perspectives to better understand PUEs A: Trade Class B: No. of Server Rack C: IT Consumption (GWh per annum) D: DC Tier Level 9
Reported Power Usage Effectiveness (PUE) A Benchmark Grade for each of the 4 Perspectives: Trade, Rack #, IT GWh & Tier-Level A) Overall Average = 2.2 PUE C) Compare with your performance E.g. You re in Trade A, your PUE=1.8 B) Sub-class Average E.g. Trade class for A = 2.0 Above Average Average Below Average Your Score: e.g. Trade Grade A Grade B Grade C Your Total Scores: Trade No. of Rack IT GWh,, & Tier Level 10
Load Factor (LF) An index to reflect how well you manage your energy usage pattern Total Consumption LF = (from 0 to ~720) Max Demand The Flatter the load, the Larger the load factor = Lower the demand charge Other Commercial & Industrial DC Load pattern 75% of DCs are enjoying discounts from High Load Factor Rider (LF > 500) 1-3% saving potential if the customers can better manage their usage pattern 11
Top Performers Insights 12
A) Temperature Setting Air Temp setting % of DC Avg PUE >= 24 deg-c 15% 1.7 >= 23 deg-c 40% 1.8 < 23 deg-c 60% 2.2 < 22 deg-c 20% 2.3 * Industry average = 22.4 deg-c * New equipment can operate at >35 deg-c Real Life Saving Examples: Case 1: 25 to 29 deg-c; 14% saving. Practically FREE of cost Case 2: 20 to 25 deg-c; 22% saving. Practically FREE of cost Considerations: Gradually increase 1 degree at a time Risk mitigation measures: (1) consult with vendor; (2) critical spares stored onsite & shortened time for fault handling; (3) more temperature sensors Hurdle to overcome: educate end-use customers; change SLA terms New trend for Colo DC: Market a DC hall as Green Zone with higher temperature setting in SLA and share the benefits with end-use customers 13
B) Cold Aisle Containment Before After Short circuit Short circuit Real Life Saving Examples: Case A: Plastic sheets to shield off cold aisle, PUE = 1.6, payback = 1 year Case B: To install high grade glass wall, expected saving 8% on chiller, payback = 4 years Considerations: Best for high power density zone where cooling demand is highly concentrated [e.g. IT density > 6kW/rack] Glass wall containment is ideal for new project. Can consider plastic sheet shield-off for retrofitting projects 14
C) Air Flow Balancing / Management Methodology: 1. Install sensors for one-off or continuous monitoring, identify hot & cold spot 2. Optimise flow rate by reducing fan speed of CRAC OR install boosters & dampers so that some CRACs can be shutdown Real Life Saving Examples: Case 1: measure bi-annually, 6% saving on reduced CRAC output, payback = 8 months Case 2: install boosters and dampers, shutdown of some CRACs, payback = 4 years Considerations: Maintenance issues for additional boosters & dampers Rebalancing on regular basis (cater IT load growth) New trend: dynamic system 15
D) Improve Load Factor IT load initiatives: Server virtualisation and switch off idle servers to reduce peak Replace energy efficient IT equipment (difference in consumption: 25%) Facility load initiatives: Case 1: Water mist spray to air-cooled evaporators, reduce 10% peak load Case 2: Optimise chiller sequencing, reduce 10% peak load Considerations: Follow regulations from WSD & EPD (Please approach CLP, welcome to share our experience with you) 16
Your Road Map to Energy Sustainability 17
Your Road Map to Energy Sustainability Your Energy Benchmarking Report A) Build awareness and keep monitoring Compare the PUE among peers in the measured variables Compare the LF among peers B) Know how to further improve Understand the Best Practices in DC industry in Hong Kong Plan your energy management initiatives in existing or new DC 18
Your Road Map to Energy Sustainability C) Talk to our dedicated account management team We provide a one-stop-shop services: 1. Energy Audit: review the energy performance of your DC and provide customised energy saving measures 2. Technical advice on major energy systems, e.g. Chillers: measure the chiller performance and plan ahead the practicality of initiatives Visit our dedicated website (clponline) to get energy efficiency best practices for Data Centres For those who haven t joined us yet: Join our DC Energy Benchmarking Service NOW! 19
CLP Your Preferred Energy Partner Please contact me if CLP can be of any assistance to your green and energy management initiatives simonlam@clp.com.hk or datacentre@clp.com.hk Thank you 20
Disclaimer CLP Power does not guarantee the accuracy of the data or the derived results or benefits from tips provided in this presentation and is not responsible for the accuracy, completeness or usefulness of the information so provided herein. It should be understood and accepted that the information contained herein may lack qualifying details and CLP will not be required to furnish any further data to support or substantiate any findings contained herein. CLP will not be liable whatsoever, in any form to whosoever, from, without limitation, any form of loss whatsoever arising or associated with the use of the contents herein. 21