India DataCenters Benchmarking
India DataCenters Benchmarking
India DataCenters Benchmarking
Acknowledgements
The authors thank NASSCOM for their efforts in coordinating data acquisition, as well as the
staff of Hewlett Packard, Maruti, Intel, Texas Instruments, and APC for kindly sharing their
facility data.
Table of Contents
Acknowledgements....................................................................................................................................... 2
Executive Summary....................................................................................................................................... 4
Introduction .................................................................................................................................................. 5
Process and Metrics ...................................................................................................................................... 5
Results ........................................................................................................................................................... 7
Conclusions ................................................................................................................................................. 17
References .................................................................................................................................................. 18
Executive Summary
Data centers constitute a large and rapidly growing sector of energy use. By one estimate, data
centers consumed 0.8% of the worlds electricity in 2005, at an aggregate cost of 7.2 billion USD
and growing by 16% each year. [1] Although the net consumption of data centers in India has yet
to be quantified, the IT-heavy nature of the countrys economy suggests relative data center
energy consumption in India may exceed the global average.
As a first step toward improving the energy performance of data centers, Lawrence Berkeley
National Laboratory has compiled a database of US data centers. In order to augment this
database, and to establish a basis for comparing US and Indian data center performance, the
USAID ECO-III project personnel solicited energy data from Indian IT companies. This paper
presents the US and Indian data, and compares the two countries data center energy
performance.
While the data set is not statistically representative of US or Indian data centers in general, it
does provide some interesting suggestions:
The Indian data centers reported performed in the midrange of their US counterparts,
perhaps slightly better on average.
There was a range of performance just like in the US, indicting opportunities for
improved efficiency; however that range was tighter in the Indian sample set.
Introduction
Data centers constitute a large and rapidly growing sector of energy use. By one estimate, data
centers consumed 0.8% of the worlds electricity in 2005, at an aggregate cost of 7.2 billion
USD. These figures are nearly double the estimates for 2000, indicating an annual growth rate of
16%. [1] Although the net consumption of data centers in India has yet to be quantified, the ITheavy nature of the countrys economy suggests relative data center energy consumption in India
exceeds the global average.
As a first step toward improving the energy performance of data centers, Lawrence Berkeley
National Laboratory (LBNL) has compiled a database of US data centers. This database allows
data center operators to benchmark their facilities against their peers, and provides data on the
energy performance of the data center stock. Each data point represents one year of energy data
from one facility, and contains metrics indicating overall energy performance, energy intensity of
air management, cooling system efficiency, and performance of the electrical power chain. In
order to augment this database, and to establish a basis for comparing US and Indian data center
performance, the USAID ECO-III project office solicited energy data from Indian IT companies.
This paper presents the US and Indian data, and compares the two countries data center energy
performance.
It should be noted that of the companies contacted, only five responded. These five data points
do not constitute a statistically representative sample of Indian data center performance, and so
the results of this analysis should be taken as suggestions, rather than strongly supported
conclusions.
ID
Data
Data Center Energy Data
dE1
Total Electrical Energy Use (kWh)
dE2
IT Electrical Energy Use (kWh)
dE3
HVAC Electrical Energy Use (kWh)
dE4
Total Fuel Energy Use (kWh)
dE5
Total District Steam Energy Use (kWh)
Total District Chilled Water Energy Use
dE6
(kWh)
Air Management
dA1
Supply Air Temperature (C)
dA2
Return Air Temperature (C)
dA3
Supply Air Relative Humidity (%)
dA4
Return Air Relative Humidity (%)
dA5
Rack Inlet Mean Temperature (C)
dA6
Rack Outlet Mean Temperature (C)
dA7
Total Fan Power (Supply and Return) (W)
Total Fan Airflow rate (Supply and Return)
dA8
(m3/s)
Name
Overall Data Center Performance Metrics
Data Center Infrastructure Efficiency (DCiE)
Power Usage Effectiveness (PUE)
HVAC System Effectiveness
Air Management Metrics
Temperature Range
Humidity Range
Return Temperature Index
Airflow Efficiency
Cooling Metrics
Cooling System Efficiency
Cooling System Sizing Factor
Air Economizer Utilization Factor
Water Economizer Utilization Factor
Electrical Power Chain Metrics
UPS Load Factor
UPS Efficiency
IT Equipment Load Density
Lighting Power Density
Unit
C
%
%
W/m3/sec
kW/ton
%
%
%
%
W/sqm
W/sqm
Results
In this section, we present the US and Indian data center metrics, and briefly discuss their
calculation and significance. Please see Tables 1 and 2 for definitions of the variables. Facilities
1 through 26 are US data centers, while 27 through 31 are the five Indian data centers currently
on record. All data is presented anonymously.
In some cases, data is missing from a significant portion of the data centers on file. These gaps
represent data that were either unreported or misreported.
Data Center End-Use Breakdown
This chart shows the end-uses of data center energy, broken into IT end-uses, HVAC, non-IT
electrical end-uses, and other. Ideally all or most of the energy goes to the IT load. In reality the
portion varies in the US data set from about one third to about three quarters. Some newer data
centers are being designed with 90+% of the power going to IT, but these high performances are
often associated with data centers in cool climates and not requiring high redundancy in their
infrastructure (e.g. 2N+). The variation in the India data (last 5 bars) is less, and the average is
slightly better (with over half the energy going to IT), but still with considerable room for
improvement. The portion of data center energy going to HVAC and electrical distribution
varies considerably in both countries, but the HVAC was always the second highest load.
10
11
12
13
14
15
16
Conclusions
Once again, the data set analyzed herein is quite small, and should not be taken as statistically
representative of Indian data centers in general. However, preliminary analysis suggests the
following points:
The Indian data centers reported performed in the midrange of their US counterparts,
perhaps slightly better on average.
There was a range of performance just like in the US, indicting opportunities for
improved efficiency; however that range was tighter in the Indian sample set.
17
For further information regarding the data center benchmarking process and potential efficiencyimproving actions, or to participate, please visit http://hightech.lbl.gov/benchmarkingguides/data.html .
References
1. Koomey, Jonathan. 2007. Estimating total power consumption by servers in the U.S. and the
world. Oakland, CA: Analytics Press. February 15. (http://enterprise.amd.com/us-en/AMDBusiness/Technology-Home/Power-Management.aspx)
18