greening the data center: the it industry's energy efficiency imperative
Post on 28-Nov-2014
4.199 Views
Preview:
DESCRIPTION
TRANSCRIPT
Greening the Data Center:
The IT Industry’s Energy
Efficiency Imperative
Larry Vertal, Senior StrategistAdvanced Micro Devices
2
Why All the Interest in Data Centers?
• Critical national and global infrastructure
• Surging demand for data processing and storage
• Rising energy intensity leads to higher cost of delivering information services
• Power and cooling limits in existing facilities
• Driving data center building boom
• Cost of electricity and supporting infrastructure now surpasses capital cost of IT equipment
• Impact on regional power grids
Building the case for Government Incentives/Regulation… HR 5646
1. Directed EPA to study energy efficient servers and data centers
2. Directs federal agencies to consider energy efficiency for server purchases
Datacenter Energy Consumption Study
Results
Electric use for
servers and
infrastructure in
US in 2005: 45
billion Kilowatt
hours
Aggregate
electrical usage
doubled from
2000 to 2005
Electric bill for
servers and
infrastructure in
US in 2005:
$2.7billion
US: Energy Consumption
AMD commissioned a two studies conducted by Jonathan Koomey, staff scientist at Lawrence Berkeley National Laboratories and consulting professor at Stanford University
Datacenter Energy Consumption Study
Results
Electric use for
servers and
infrastructure in
Worldwide in
2005: 120 billion
Kilowatt hours
Aggregate
electrical usage
doubled
worldwide from
2000 to 2005
Electric bill for
servers and
infrastructure
Worldwide in
2005: $7.2billion
A Global Concern: Energy Consumption
AMD commissioned a study conducted by Jonathan Koomey, staff scientist at Lawrence Berkeley National Laboratories and consulting professor at Stanford University
The Growing Energy Challenge
ESTIMATING TOTAL POWER CONSUMPTION BY SERVERS IN THE U.S. AND THE WORLDJonathan G. Koomey, Ph.D., Staff Scientist, Lawrence Berkeley National Laboratory and
Consulting Professor, Stanford Universityhttp://www.amd.com/us-en/Corporate/VirtualPressRoom/0,,51_104_543~115850,00.html
Servers account for close to 1% of US electrical consumption in 2005
US datacenters consume more power than five 1000 megawatt nuclear power plants – at a cost of almost $3 billion
Aggregate electrical usage doubledfrom 60 billion kWh/yr in 2000 to 120 billion in 2005 worldwide
Electric bill for servers and infrastructure worldwide in 2005: $7.3 billion
Annual US data center power consumption could power Las Vegas for one year*
*http://www.nevadapower.com/company/facts/
5
Source: US EPA Data Center Study
Comparison of Projected Electricity Use
by US Data Centers 2007-2011
6
7
Energy Independence and Security Act of 2007
Produce specifications, measurements, best practices, and benchmarks for data centers with regards to:
• Performance and use of servers, data storage devices,
and other information technology equipment
• Efficiency of heating, ventilation, and air conditioning, cooling,
and power conditioning systems
• Energy savings from the adoption of software and data
management techniques
Passed by Congress & signed by President – December 2007
1. National Program
2. Establishes program in cooperation between EPA, Dept. of Energy & an “ Industry Data Center Efficiency Organization “
Datacenter Power Mix
What’s Consuming all this Power?
Lighting power consumption1% - 2%
Server Power Consumption Impacts Power throughout the Datacenter
Computer Room Air Conditioner power
consumption
23% - 54%
Battery Backup power consumption
6% - 13%
Server, network and storage power consumption
38% - 63%
9
What’s Consuming all this Power?
For every 3 watts into the datacenter,only 1 watt actually powers the device!
~2 WattsCooling!
1 WattPowers the server or workstation
200 mWPower distribution and electrical power supply
10
17%
12%
38%
21%
12% Power Consumption
Cooling
Both Equally
Important
Suspect That Both
Are Issues But Do
Not Track At This
Time
Neither Presents An
Issue At This Time
Data Center Power Consumption And Cooling
Issues
Base: Total Respondents (1,177)
Q. Between the two, what is the primary issue you face in your data center today?
Power Consumption/Cooling Issues
Tracked By Company (Net): 71%
15%
16%
23%
25%
26%
44%Increased Amount Of Power Supplied To
Data Center
Stopped Buying More Servers And/Or
Consolidated Existing Equipment
Implemented "Cool Aisle/Hot Aisle"
Layout
Increased Size Of Data Center
Other
None Of The Above
How Addressed These Data Center Issues
Base: Power Consumption/Cooling Issues Tracked By Company (834)
Q. How has your company addressed this issue to date?
85% Have Addressed Data Center Issues
Yes
21%
Don't Know
10%
No
69%
Rack Space Go Unused
Impact on Rack Space Usage
Base: Total Respondents (1,177)
Q. Does rack space in your data center go unused due to
power consumption and cooling issues?
Over 50%
1%
Under 10%
18%
21%-50%
20%
10%-20%
61%
Average Rack Space Unused: 18%
Base: Rack Space Goes Unused (250)
Q. What percentage of rack space goes unused
due to power consumption and cooling issues?
Yes
29%Don't Know
15%
No
56%
Data Center Power Consumption And
Cooling A Way To Lower Costs
Base: Total Respondents (1,177)
Q. Is data center power consumption and cooling being investigated by business management as a way to lower
operational costs?
Yes, Affected
29%Don't Know
9%
Not Affected
62%
Power Consumption And Cooling Issues Affected Server
Purchase Decision In Past 12 Months
Base: Total Respondents (1,177)
Q. Have power consumption and cooling issues
affected a server purchase decision in the past
12 months?
7%
25%
34%
40%
64%Used Server Vendor's Published
Specs
Used Server's Power Supply Rating
Conducted Power/Cooling Tests In-
House
Used Server Vendor's Power Budget
“Calculators”
Other
Base: Issues Affected Server Purchase Decision (336)
Q. How did you determine whether the chosen server met
your criteria?
Server Purchase Decision CriteriaAffected Server Purchase Decision
Base: Would Implement System Level Tools (834)Q. If system-level tools were available that would allow you to manage
power and cooling at the server rack level, would you implement
them? Which management package would you like to see these
tools integrate with?
Implementation of System-Level Tools To Manage
Power And Cooling
31%
9%
5%
9%
17%
29%HP Openview
IBM Tivoli
LANDesk
CA Unicenter
Other
Don't Know
Management Package Would Like To Integrate
71% Would Implement System-Level Tools 29% Would Not Implement System-Level Tools
Reasons Would Not Integrate
24%
20%
38%
52%
Don't See Benefit
Of Managing
Down To That
Level
Not Enough IT
Staff To Track
Another
Management Tool
Consider This An
Issue That Server
Vendors Should
Resolve
Other
Base: Would Not Implement System Level Tools (343)
Q. If no, why not?
Base: Total Respondents (1,177)
Q. What type of processor powers the majority [75% or more] of the servers in your data center?
Q. What type of processor powers the balance of the servers in your data center?
Types Of Processors In Data Center
8%
4%
4%
8%
9%
67%Intel Xeon
IBM POWER
Sun UltraSPARC
AMD Opteron
Other
Don't Know 25%
11%
15%
20%
23%
24%Sun UltraSPARC
Intel Xeon
IBM POWER
AMD Opteron
Other
None, All Servers Are Powered By
The Same Processor Type
Processors Powering The Majority Of
Servers In Data Center
Processors Powering The Balance Of
Servers In Data Center
14%
14%
17%
25%
31%2U-3U
1U
4U-5U
Blade
Other
Server Form Factors Currently Installed(Average Percentage of Servers in Data Center)
Base: Total Respondents (1,177)
Q. What server form factors do you currently have installed in your data center (breakdown by percentage)?
19
Infrastructure Impact of Quad-Core
Datacenter-Level
Wattage based on 2P systems, 8 DIMMs, TDP wattage for ‘Dempsey’, ‘Woodcrest’ & ‘Clovertown’ is estimated based on current publicly available processor and chipset values, AMD estimates, and an incremental 100watts for fans, storage, and power supply. (see, eg: http://techreport.com/etc/2006q2/woodcrest/index.x?pg=2 and is subject to change. The examples contained herein are intended for informational purposes only. Other factors will affect real-world power consumption.
Datacenter Power Budget7 Megawatts
Increased Power Increased Power
and Thermals can and Thermals can
result in result in
expensive facility expensive facility
upgrades:upgrades:
$ facility build$ facility build--outout
$ electricity cost $ electricity cost
$ power distribution$ power distribution
$ HVAC equipment$ HVAC equipment
14% Wasted 14% Wasted
Space with Space with
QuadQuad--Core Core
XeonXeon
10% Wasted 10% Wasted
Space with Space with
DualDual--Core Core
XeonXeon
Example: SecondExample: Second--
Generation Generation
AMD AMD OpteronOpteron™™
processors with processors with
upgrade path to upgrade path to
quadquad--core within core within
existing power & existing power &
thermal thermal
envelopesenvelopes
Dual-core
100 Racks
8000 total cores
Quad-core
100 Racks
16,000 total cores
Dual-core
90 Racks
6840 total cores
Quad-core
85 Racks
12,240 total cores
Transition toQuad-core
DifficultTransition toQuad-core
POWERLIMITED
POWERLIMITED
20
One “Brain” to Four “Brains”:
Greater computing but same power & thermals
Greater
Performan
ce
Same Po
wer
25% better performance
30% better performance/watt
Example: Example:
SecondSecond--
Generation Generation
AMD AMD OpteronOpteron™™
processors processors
with upgrade with upgrade
path to quadpath to quad--
core within core within
existing power existing power
& thermal & thermal
envelopesenvelopes
Performance-Per-Watt Scalability
21
Putting Quad-Core to Work
1 Core 2 Cores 4 Cores Multi-Cores
Optimized Software
Un-optimized Software
Software need to be properly coded to take advantage of multiple cores
Development tools can help create multi-threaded applications− Compilers
− Java
− MSFT .NET Framework
− Libraries
Perf
orm
ance
Core Count
Silicon drives performance gains
Software drives performance gains
Multi-Threaded Software Considerations
Launched in February 2007, today more than 158 member companies
www.thegreengrid.org
22
In 2006, AMD partnered with IT industry leaders to create
The Green Grid consortium
• A global consortium dedicated to developing models, metrics and standards and to promoting energy efficiency for data centers and information services delivery
• Founding Members: AMD, APC, Dell, HP, IBM, Intel, Microsoft, Rackable Systems, SprayCool, Sun Microsystems and VMware (EMC)
The Green Grid
Understanding the Issue
• Despite the fact that power consumption is one of the most important issues facing IT today, there is a lack of guidelines and resources available for those looking to drive a change:
– Prior to the formation of The Green Grid, there were few standards efforts for measuring energy efficiency in the data center
– Most information on data center efficiency is departmentalized and proprietary
– No clear roadmap for future data center design and operations; IT managers need information on what to start implementing now
Technical Work Groups
• Data Collection and Analysis– Investigations and analysis into data center efficiency
problem space
• Data Center Technology and Strategy– Existing and emergent technologies for data center
efficiency
• Data Center Operations
– Use models, operational strategies and best practices
– Equipment standards
• Data Center Metrics and Measurements– Data center characteristics and performance metrics
– Measurement protocols for data collection
Announced at February
the Technical Forum
• Data Center Rating System – industry discussion on data center ratings with a number of parties – including California Energy Commission, DOE, Electric Power Research Institute, EPA, European Commission and The Green Grid – who have started working on data center energy efficiency rating systems
• Data Center Productivity – session highlighting The Green Grid’s progress in developing metrics for data centers that relate useful work performed within data centers to the resources required to complete that work
• Measuring DCiE Today – panel discussion on what various companies are doing with regard to monitoring and measuring DCiE and PUE in their daily operations
Delivering on the Promise
The Green Grid is the industry voice on energy efficiency for data centers and business computing ecosystems
• Defining meaningful, user-centric models and metrics
• Developing standards, measurement methods, best practices and technologies to improve performance against the defined metrics
• Promoting the adoption of energy-efficient standards, processes, measurements and technologies
THANK YOU
top related