SM
SM
Power Usage Effectiveness (PUE)
& Data Center Infrastructure
Efficiency (DCiE) Progress
Dan Azevedo, Symantec Jon Haas, Intel
Jud Cooley, Sun Microsystems, Inc. Mark Monroe, Sun Microsystems, Inc.
SM
SM
Agenda
•
PUE / DCiE Background & Next Steps
•
DCiE Detailed Analysis
•
PUE / DCiE Usage & Public Reporting Guidelines
•
PUE / DCiE Calculation Tool
•
Free Cooling Map
SM
SM
PUE / DCiE Background & Next
Steps
SM
SM
DCiE Detailed Analysis
Dan Azevedo, Symantec
SM
SM
Executive Summary
The DCiE Detailed Analysis is to support and
strengthen the Data Center infrastructure
Efficiency (DCiE) metric. The intent is to enable
data center managers collectively to derive and
use DCiE as a global standard metric.
SM
SM
DCiE Detailed Analysis
This White Paper represents the first in a series that will
provide a more detailed analysis of the PUE / DCiE
Metric and represents the “Basic” level in the below table
with the assumption of a dedicated data center facility:
Level 1 (Basic)
Level 2
(Intermediate) Level 3 (Advanced)
IT Equipment Power Measurement From…
UPS PDU Server,….
Total Facility Power Measurement From…
Data Center input power
Data Center input less shared HVAC
Data Center input less shared HVAC plus building lighting, security…
Minimum Measurement Interval
Monthly/Weekly Daily Continuous (xx min)
SM
SM
SM
SM
SM
SM
DCiE Detailed Analysis
Table of Contents
Introduction
Type of Data Center
Background and Assumptions
Climate and Location
DCiE Overview
Data Center Design and Operation
DCiE Subcomponents
Measurement Period
Obtaining Required Data
Sometimes DCiE Seems To Go The
Wrong Way
Estimating Power Consumption
Driving DCiE Efficiency
Measuring Power Consumption
DCiE and Future Technology
Advancements
Power Metering Requirements
Multiuse Facilities
Factors Affecting DCiE
Summary
SM
SM
Summary
This paper provides:
•
Prescriptive information on how to calculate the
efficiency metrics
•
Multiple levels of implementation which enables a
very simple “basic” approach to as much depth and
complexity as desire
•
Equipment categorization: infrastructure vs. IT
equipment
•
Power metering requirements
•
Guidance on interpreting the metrics
SM
SM
Usage And Public Reporting Guidelines for
The Green Grid’s Infrastructure Metrics
Jon Haas, Intel
SM
SM
Reasons for Guidelines
•
PUE/DCiE have become commonly used metrics for data center efficiency
•
Factors such as environmental conditions, a data center’s physical
characteristics, and timing and frequency of measurements can play a
significant role in final results
– In addition, some individuals and organizations have used these metrics to promote design ideas that, while valuable, can provide misleading results
•
Standard guidelines for the use and reporting of these metrics will:
– Enable a common language and nomenclature to compare results
– provide a process to support accurate, consistent results
– Provide additional transparency and accountabilitly in the process of measurement for public claims
•
In addition, these guidelines will support The Green Grid in:
– develop an auditing/certification process for the industry around these metrics
– develop a recognition program for those organizations showing strong results and/or improvements in data center energy efficiency
SM
SM
Publicly reporting your Data
•
The Green Grid will provide a means for organizations to publicly report their
measurements
– record their PUE/DCiE measurement data to The Green Grid
– provide background information about the data center, reporting the results and the manner in which, and conditions under which, that data was collected
– provide a means to report any issues or difficulties in following TGG’s proscribe processes and guidelines
•
The Green Grid will also provide a portal through which organizations can
view summary details on publicly reported PUE and DCiE results
– This portal will provide summary data on all registered reported results
● All reported results meeting TGG guidelines will be given a registration number that can be referenced
● A subset of provided data will be publicly visible (see appendix 2 of WP or consent form)
– TGG will offer members the opportunity to provide a link to additional information about their result or data center via additional optional reporting
– Organizations reporting results have the option to participate in TGG awards or recognition programs by complying with additional requirements
SM
SM
Classes of Measurement Results
Class Description Benefit to Reporting Organization
Unrecognized A publicly reported result with no claims of following TGG’s guidelines. TGG will not comment on unrecognized results.
Reported A publicly reported result by the reporting organization claiming they followed TGG’s measurement
recommendations and nomenclature guidelines. TGG will not comment on Reported results.
Reporting organization can use standard materials from The Green Grid to explain process and results to audience.
Registered A publicly reported result, with key
report contextual data provided to TGG by the reporting organization to TGG’s data center performance database.
Official registration of
reported result. Receipt of a registration number from TGG. Link to public report data from TGG’s website.
Certified A publicly reported result, with key additional data required for third-party validation or certification of results, provided to TGG by the reporting organization.
All benefits applicable to registered results, plus, consideration of reported
results in future TGG award or recognition programs.
SM
SM
Usage and Reporting
Guidelines
To receive a PUE or DCiE registration number from The Green Grid:
– Follow The Green Grid’s measurement guidelines in the most recent revision of “DCiE Detailed Analysis”
– Report the process through which the data was collected (L1, L2, L3) using the nomenclature in the most recent revision of “Usage and Public Reporting Guidelines for The Green Grid’s Infrastructure Metrics”
– Report your measurement data and input the required information into the Green Grid’s Metrics Data Collection Database (www.thegreengrid.org)
– Agree to The Green Grid’s metrics public reporting consent form
– If requesting to be considered for a TGG award, provide the extended data set information into the Green Grid’s Data Collection Database for
certified measurements
–
Additionally
–
Report to The Green Grid any issues or difficulties in following TGG’s
proscribed processes and guidelines
– (optional, ‘co-promotion’) send the PUE or DCiE registration number, along with a link to any original source material or publication, to TGG at
SM
SM
PUE/DCiE Nomenclature
Structure
•
All PUE and DCiE measurements should be reported with subscripts that
identify:
– The accuracy level of the measurement
● See “The Green Grid’s Metrics: Data Center Infrastructure Efficiency Detailed Analysis”
– The timeframe of the measurement
● E.g. yearly, monthly, weekly, daily
– The granularity of the measurement
● Frequency with which the data was collected
● E.g., monthly, weekly, daily, ‘continuous’
•
Official PUE/DCiE reports will provide these information with subscripts
appended to the Metric
– i.e. PUEx,y or DCiEx,y, where
– ‘x’ describes the ‘Level’ of the measurement (L1, L2, L3)
● L1, L2, and L3 are defined in the TGG White Paper: “The Green Grid Metrics: Data Center Infrastructure Efficiency (DCiE) Detailed Analysis”
– ‘y’ describes the ‘Timeframe’ and ‘Granularity’ of the measurement
SM
SM
Measurement Accuracy Reporting
(from DCiE Detailed Analysis WP)
Level 1 (Basic) Level 2 (Intermediate) Level 3 (Advanced) IT Equipment Power Measurement From…
UPS PDU Server,….
Total Facility Power Measurement From…
Data Center input power
Data Center input less shared HVAC
Data Center input less shared HVAC plus building lighting, security…
Minimum Measurement Interval
Monthly/Weekly Daily Continuous (xx min)
SM
SM
Measurement Timescale and
Granularity Reporting
•
The ‘y’ subscript on PUE or DCiE denotes the ‘Timeframe’ and
‘Granularity’ of the measurement
–
‘y’ = timeframe identifier + frequency identifier [‘M’, ’W’, ’D’, ‘C’]
•
Frequency identifier:
–
‘M’ for measurements taken monthly
–
‘W’ for measurements taken weekly
–
‘D’ for measurements taken daily
–
‘C’ for measurements taken ‘continuously’ (with frequency one hour or
less)
•
Except for those measurements collected with ‘continuous’ data, all
data points that are averaged in the final measurement must be
taken at the same time of day
SM
SM
Example Reported Results
0.45 DCiEL1,-- Single DCiE measurement (0.45) taken using a Level 1 meter placement
0.51 DCiEL1,YM Yearly average DCiE (0.51), using data points gathered monthly with a
Level 1 meter placement
1.6 PUEL1,MW Monthly average PUE (1.6) using data points gathered weekly with a Level
1 meter placement
0.43DCiEL1,WD Weekly average DCiE (0.43), using data points gathered daily with a Level
1 meter placement
1.8 PUEL2,WC Weekly average PUE using data points gathered continuously with a Level
2 meter placement.
2.1 PUEL3,YC Yearly average PUE (2.1) using continuous measurements with a Level 3
SM
SM
Specific Reporting Issues
•
PUEs of < 1.0 and DCiEs of > 1.0 are *not* allowed
•
Elements that go into the calculation
–
Power consumed by IT equipment
–
Power lost in distribution
–
Power consumed by cooling equipment
–
These elements are always
≥ 0
•
Accounting for re-use of waste heat
–
Waste heat re-use *does not* get included in either PUE or DCiE
–
TGG will look into creating an additional metric to cover this
•
Accounting for power generated on site
–
PUE / DCiE is meant to cover areas of energy use in the data center
–
Generation of power on-site is not incorporated into either the power
distribution losses or cooling equipment requirements
SM
SM
Conclusion
•
We are driving the right behaviour!
The Green Grid Metrics for data center efficiency are
gaining widespread adoption, use and visibility
Organizations are publicly reporting their
measurements and claiming highly efficient operations
•
The Green Grid is implementing public reporting
guidelines that includes standard processes and
nomenclature to make the measured results
more transparent and consistent
•
Stay away from PUEs of < 1.0 and DCiEs of >
SM
SM
PUE Calculation Tool
Jud Cooley, Sun Microsystems, Inc.
SM
SM
Rationale and Objectives
•
Sponsor: Container Assessment Task Force to allow case
study comparisons
•
Easily and accurately compute PUE
–
In a consistent manner
–
Allowing for apples-apples comparisons
–
Container boundaries
–
Building or room boundaries
•
Web based
–
One sheet printed report
SM
SM
PUE with Physical boundaries
•
Management reporting to visualize overhead
Power: Transfer Switch, UPS, Transformer, Generator, Lighting, etc.
Cooling: Tower, Condenser, Chiller, Pumps, etc.
Other: Fire Supression, Control, Security, etc.
IT Equipment: Servers, etc.
Facility Boundary
Container or Room Boundary
Facility equipment: air movers, power
distribution, UPS, etc.
SM
SM
Web screen shot example
(Spreadsheet Prototype, artificial data)
Seq Class Type Detail
1 Facility Power Transfer Switch
2 Facility Power UPS 5.0
3 Facility Power DC Batteries / Rectifiers (non UPS – Telco Nodes) 4 Facility Power Generator
5 Facility Power Transformer (step down) 5.0 13 Facility HVAC Chillers 25.0
19 Facility HVAC Return Fans 2.0
...
30 IT Equipment Compute Devices Servers 50.0 31 IT Equipment Network Devices Switches
32 IT Equipment Network Devices Routers
40 IT Equipment Other Other 2.0 2.0
...
Total Facility + IT Power 32.0 59.0 Total IT Power 2.0 52.0 Total Facility Power (kW) 91.0
Total IT Power 54.0
PUE 1.69
DciE 0.59
Total Power (kW) at Container Boundary 59.0 Total IT Power in Container 52.0 Partial PUE (Container Boundary) 1.13 Partial DciE (Container Boundary) 0.88
Facility (kW)
Container (kW)
SM
SM
Output
•
Single page PDF output
–
Fileable report identifying configuration, date, data
–
Useable for direct apples-apples comparisons
•
Data saved in csv or a Data Base format
–
TBD based on development discussion
–
Consistent with other Green Grid tools
–
Importable into spreadsheet
SM
SM
Controversy over air movers
•
Fan-less servers with rack fans rely on those fans
•
Rack-based air movers might be considered either IT or
facility depending on your point of view
•
Proposed position is that the air mover power is part of
the IT equipment
•
Needs to be discussed in Committees
PUE Calculator Tool will allow a separate line to enter air
mover power that is not included inside the physical
boundary of the server, yet is required for the IT
operation (to expose the ambiguity). That power will be
allocated to IT, but will be left visible in the tool and the
output.
SM
SM
Status and Timing
•
Consultants hired
•
In development now
•
Alpha version to be tested by Green Grid Task Force
members in Q1 2009
SM
SM
Free Cooling Map/Web Tool
Mark Monroe, Sun Microsystems, Inc.
SM
SM
Concept and Features
•
Estimate the number of hours of free cooling that may
be possible based on weather data at any location in
the U.S.
–
Consider European and APAC data in the design so that data
can be added in the future without major changes to the
application or database.
•
Climate data will be analyzed to determine the number
of hours per year that either fresh air or evaporative
cooling might be used in a data center.
SM
SM
Concept and Features (cont’d)
•
Two colorful maps must be produced;
–
one showing the U.S. color-coded to show the number of hours
of fresh air cooling possible,
–
the second map will show the U.S. color-coded to show the
number of hours of evaporative cooling possible.
SM
SM
SM
SM
Concept and Features (cont’d)
•
A web tool must be produced which allows a
user to enter their own location, via zip code,
and the data center set points for temperature
and humidity that they which to examine for
hours of free cooling
SM
SM
SM
RFP Responses
•
weatherbank.com selected to implement
•
Response within budget estimates
•
Approved by Board 21 Jan 09
•
Prototype ready now
–
Seek approval from TC and Board in Feb
SM
SM
Roadmap/Futures
•
Individual maps available for $150 direct from
weatherbank.com
•
Feedback page allows users to consult directly
•
Review in 3 months, determine enhancements
•
Possibilities:
–
European and APAC data/locations/calculations
–
Simple model for water used by evaporative cooling
–
Input location by latitude/longitude
–
Redisplay map with user-input parameters
–
Connect with DSIRE and other incentive databases
SM
SM
Power Distribution Configurations
Online Efficiency Estimator Tool
SM
SM
Tool Purpose
The tool:
•
Is for use as a companion to
–
“Qualitative Analysis of Power Distribution Configurations for
Data Centers”, TGG WP #4 and
–
“Quantitative Efficiency Analysis of Power Distribution
Configurations for Data Centers”, TGG WP #16
•
Reduces the need for paper rewrites, as the tool can be
easily updated regularly
•
Allows readers and users to test their own components
and proposed system architectures
SM
SM
Features
•
Users are able to:
–
Compare two topologies at a time
–
Input data points (load vs efficiency) of their own components
for the calculations
–
Create their own topologies based on their components or
pre-defined TGG components
–
Save components and topologies
–
Print input and results in PDF
•
Tool will automatically calculate full efficiency curves for
–
components
SM
SM
Tool
SM
SM
Availability
•
The tool is scheduled to be available around Q2
2009
•
It will be accessible to members only via The
SM
SM