2. 2
What is Information
Technology (IT)?
Information
+ Technology
Technology: fire, wheel
Early information technology: Dewy
decimal system
Information collections continue to expand
Compiling data into information became
exponentially labor intensive
3. 3
Information Systems
Information
systems make it possible to
process large amounts of data into
information
Now that we have good information how
do we quickly share research findings?
Networking; the internet
4. 4
Ubiquitous Communication
The
internet gave rise to ubiquitous realtime communications
Anyone can transmit:
Files
Photos
Video
We
use the internet to connect with
others
5. 5
Convenience
Bank
by phone
Turn by turn directions
Researching purchases
Home buying
Meal planning
Home automation
Smart Grid
9. 9
Smart Grid & IT
Information
technology and the
power grid are interdependent
2003 North East Blackout
The extent and duration of the
outage was due to a race
condition in the SCADA system
10. 10
Smart Grid & IT
Smart
Grid expands this interdependence
However it also enhances the reliability and
redundancy of the SCADA systems
More situations are automated to avoid
human error
11. 11
Data Centers & Smart Grid
Can
coordinate energy consumption
during
Low periods
Periods where a high percentage of
renewable energy such as wind or solar is
available
12. 12
What types of IT processes
could take advantage of this?
Batch
jobs such as running
Reports
Back ups
Scheduled maintenance
13. 13
Big Data Problem
All
of these integrated systems will create
vast amounts of data
This data will be used for modeling to make
real time decisions
This data will depend upon information
systems infrastructure to maintain; creating
additional IT processing and archive loads.
14. 14
What’s the big deal storage
space is cheap?
Not
really
Hard
drive usage must be powered
Additional load for backups
Can be a huge deal in a virtualized
environment
Memory usage is also decreased by 34%
Multiply that by 100 in a virtual environment, the
resource savings is huge
15. 15
Storage Problem
Vast
amounts of data some will be used in
real-time situational analysis to allow
Independent Systems Operators to make
operational decisions.
Past data may not be accessed daily, but
will need to be readily available for
generation of reports to make strategic
decisions
Accessibility to large amounts of data
creates substantial energy overhead
16. 16
Energy Efficient Data Storage
Hard
Disk Drive (HDD) vs. Solid State Drives
(SDD)
HDD
Uses 80% of maximum power draw at start-up
SDD
Linear proportion usage to power relationship
This
means that any HDD active, but running
at under 80% utilization is wasting energy.
17. 17
Energy Efficient Data Storage
So
is the answer to use all solid state
drives?
That is one possible answer, but it is currently
an expensive answer.
Other
possibilities:
Storage consolidation
Virtualization
is one possible method of
achieving this goal
Shared storage arrays are another
18. 18
Storage Problem
Increased use of
technologies such
as memcached for
frequently
accessed data
This technique uses RAM
caching
Used in conjunction with
compression this can be
very useful for
transactional loads
Hybrid SDD/HDD
storage systems
managed
intelligently
19. 19
Data Center Power
consumption
In
2005 data center power usage was 1%
of the worlds power consumption.
The current power usage is estimated to be as
high as 1.5% world wide and up to 2.2% of US
power consumption.
20. 20
Data Center Power
consumption
Why
should you care?
Every year there are rolling blackouts though
the summer because utilities cannot keep up
with demand.
Energy costs are higher than equipment costs.
Loss of electricity or computing capabilities can
pose national security risks
21. 21
Efficiency Basics
Power
Usage Effectiveness (PUE)
developed by the Green Grid is a widely
accepted measure of data center
efficiency.
PUE Calculation
Total data center energy consumption
IT energy consumption=PUE
22. 22
PUE explained
A
PUE of one means that all energy
consumptions is being used by the
servers, storage devices and networking
equipment.
Reaching one is the goal, but may not be
possible currently…
However, Google has reached reach a Power
Usage Effectiveness (PUE) of 1.16.
23. 23
There is a better way…
According to Google “if all data
centers operated at the same
efficiency as ours, the U.S. alone
would save enough electricity to
power every household within the
city limits of Atlanta, Los Angeles,
Chicago, and Washington, D.C.”
24. 24
Google
Google tracks its PUE
Data centers are run at 80 degree Fahrenheit
or more
This “conforms with both the American Society of
Heating, Refrigerating, and Air Conditioning Engineers'
recommendations and most IT equipment
manufacturers' specs.”
Data centers can be cooled without chillers
Remaining chillers have disabled “dehumidifying and
reheating functions on CRACs. Most are set to
dehumidify air to 40% and reheat air if the return air is too
cold, but these functions aren’t needed.”
Each data center element is designed to
operate at optimal efficiency
25. 25
Google: How did they do it?
Power Supply
The average power supply converts power from
AC to DC accounting for 30 to 40% in power
loss.
This process also produces heat
Google power supplies do not perform this
costly conversion and have an integrated
UPS.
“We’ve also cut out 2 of the AC / DC
conversion stages by putting back-up batteries
on the server racks themselves.”
26. 26
Google: Custom Servers
Parts
are omitted “on servers that aren't
needed for our applications. Hardware is
limited to what is necessary for the
applications to run, and does not include
unnecessary components such as
peripheral connectors or video cards. We
also optimize our servers and racks to use
minimal fan power, and the fans are
controlled to spin only as fast as necessary
to keep the server temperature below a
threshold.”
28. 28
Facebook
Facebook
recently adopted a novel
power distribution design that
removes uninterruptible power supply
(UPS) and power distribution units (PDUs)
from the data center.
The
new design shifts the UPS and battery
backup functions from the data center
by adding 12 volt battery cabinets
Facebook’s
datacenter has a PUE of 1.07
29. 29
Role of Software
”(T)he
only way to have software
consume zero resources is not to run it at
all. Even running very well-behaved
software at the minimum will, in
practice, require some resource overhead.”
Virtualization
Power Management
Node Management
Scheduling
Lean Operating systems
Software development
De-duplication
30. 30
Virtualization
Virtualization
is useful for consolidation of
under utilized servers
DO: consolidate servers that typically have
usage under 70%
DO NOT: consolidate servers that have
usage higher than 90%
Unless
the server resources are more than
enough to cover usage patterns including the
other virtualized servers consolidated into the
new server
WARNING:
Consolidating several high
usage servers on one server may be a
recipe for disaster.
31. 31
Virtualization
Remember:
Virtualization can lead to the
need to rework cooling in the data
center.
Fewer machines could lead to
unnecessary cooling of space no longer
occupied by servers. This could result in
wasted cooling.
32. 32
Lean Operating systems
Graphical
efficient
1 watt can be saved at the server level
though using a command line only server
operating system
The overhead of the GUI is about 100
threads
This does not mean no Windows server OS
Windows
User Interfaces (GUI) are not
server 2008 R2 Core
Uses less disk space
Standard
7.5 GB
Core 3.5 GB
33. 33
Software Development
CPU
UTILIZATION
Write event-triggered not time-based checks
Use batch processing for processes that cannot
be event based
Make sure batch jobs can be adjusted for time
conflicts
MEMORY
UTILIZATION
Avoid memory leaks
Release memory when it is no longer
needed, don’t wait for the system to do it foru
you.
34. 34
Software Development
I/O
UTILIZATION
Buffer/batch I/O requests
EFFICIENT
SYSTEM STACK
“Features to reduce power consumption of
underutilized system resources have become
pervasive in even the largest systems, and the
software layers responsible for managing those
resources must evolve in turn”
Implement policies:
Increase performance of resources in use
Reduce power consumption for resources not
in use
35. 35
Software Development
(Continued)
“Resource consumers clearly have a significant
opportunity either to contribute to or undermine
the efficiency of the broader stack. Though
getting programmers to think differently about the
way they design software is more than a technical
problem, tools such as PowerTOP represent a
great first step by providing programmers and
administrators with observability into software
inefficiency, a point of reference for
optimization, and awareness of the important role
software plays in energy-efficient computing."
36. 36
Hardware
What
is better one really powerful server
of several low powered servers?
Answer:
It depends.
Many low powered servers are ideal for
transactional loads
Powerful servers are better for simulation and
computation
37. 37
Conclusion
"Any
intelligent fool can make things
bigger, more complex, and more violent.
It takes a touch of genius -- and a lot of
courage -- to move in the opposite
direction.” ~Albert Einstein
In
IT this kind of thinking results in huge
payoffs when implemented correctly.
Notas do Editor
Point out that using the internet to do research saves money and fossil fuels
http://www.research.att.com/export/sites/att_labs/library/image_gallery/articles/2010_01/2010001_smart_main_dk2b.jpgTalk about how smart grid products same money and increase convenience.