good practices in the use of ict equipment for electricity savings at a university campus
DESCRIPTION
This presentation has been presented at the 5th International Green Computing Conference (IGCC), Dallas, Texas USA, November 2014. Abstract of the corresponding research paper follows: The commercial sector is responsible in a large degree for the overall energy consumption around the world and Information and Communication Technologies (ICT) constitute an important category of electricity loads which is becoming dominant in offices. Recognizing the importance of using ICT equipment more rationally for saving energy in commercial buildings, we have performed an energy audit of ICT loads at the School of Design and Environment, National University of Singapore, using smart power outlets. After a six-month period through which we have measured and analyzed the consumption of the most representative ICT devices available at the school, we have identified various good practices in the use of office equipment, in order to reduce the total electricity footprint of the building. Most of our practices can be generalized for offices and commercial buildings. Cite the paper as follows: ndreas Kamilaris, Dang Truong Hoang Ngan, Alexandros Pantazaras, Balaji Kalluri, Sekhar Kondepudi and Tham Kwok Wai. Good Practices in the Use of ICT Equipment for Electricity Savings at a University Campus. In Proc. of the 5th International Green Computing Conference (IGCC), Dallas, Texas USA, November 2014.TRANSCRIPT
Recommended Practices in the Use of ICT Equipment for Energy Savings
Andreas Kamilaris, Dang Truong Hoang Ngan, Alexandros Pantazaras, Balaji Kalluri, Sekhar Kondepudi and Tham Kwok Wai
Dallas, TX : Nov 3-5 2014
Green ICT :
Importance of ICT Loads
2 © 2014 Sekhar Kondepudi
MELs (Miscellaneous Electric Loads)
MELs account for more than 20% of the primary energy used in commercial buildings and this is expected to rise to 40% by 2035.
MELs are transforming into dominant electrical loads
As can be seen the electrical loads by traditional uses is expected to decline, while the same for MELs is increasing rapidly.
3 © 2014 Sekhar Kondepudi
Motivation
• A study measuring consumption of MELs in a controlled environment in University of California in San Diego revealed that MELs category of ICT equipment accounts for:
–70% of the electrical loads during peak hours
–80% of the electrical loads during off-peak hours
• Therefore energy audit focusing on ICT equipment is very important, when minimizing the energy foot-print of a building.
4 © 2014 Sekhar Kondepudi
Questions - Typical ICT Devices in an Office Building
• What are the different power (plug load) behaviors of different ICT devices ?• What are the different ICT device classes ?• Can these plug loads and their related parameters be characterized ?• What is the transient behavior of these devices – moving from one state to
another?• Are there similar patterns within a device class?• How to manage increasing ICT loads in next-gen smart buildings ?• Can we develop Predictive Algorithms for Disaggregating Multiple Loads from a
Single Measurement
5 © 2014 Sekhar Kondepudi
Goal & Approach
• Goal–To develop Best Practices on the Use of ICT Devices in the University Campus using the School of Design & Environment (SDE) as a proxy (3 Departments, 3 Buildings)
• Quantitative–Detailed Inventory of ICT Devices
–Measurement of Individual ICT Devices
–Field Measurements & Monitoring
• Qualitative–Surveys and interviews with users and facility managers
6 © 2014 Sekhar Kondepudi
Inventory : 225 rooms with 1,534 ICT devices
• Desktop PCs + Displays = 1300
• Imaging (Printers, Scanners, MFD) = 150
• VoIP Phones = 85
• Projectors = 44;
• Faculty offices = 108, predominantly in SDE1-L4&L5 then in SDE3-L2&L3
• Admin Staff = 104, predominantly in SDE1-L3&L5 then in SDE2-L1&L2
• Limited Laptops – Not permanently attached ( only 5 )
• Desktops: 41%• Display Monitors: 41%• Imaging equipment: 8%• VoIP, projectors, external LCD screens:10%
7 © 2014 Sekhar Kondepudi
Measurements
• Energy Metering HardwareEmployed smart power outlets or smart plugs to sandwiched in the middle of the socket and electrical appliance plug to measure their consumption. • These motes sample the current, voltage, active and apparent power of the
attached load.
• They include a low-power processor, radio and integrated antenna.
• Developed drivers in Java for parsing the measurements and storing them in a database for statistical analysis, similar to smart homes concept.
8 © 2014 Sekhar Kondepudi
ICT Devices Sample Test Matrix
Monitor Projector Printer Desktop Laptop
ON-Low Brightness OFF OFF OFF OFF
ON-Med Brightness SLEEP OFF – BOOT UP – IDLE OFF to ON OFF to ON
ON-High Brightness ON IDLE – PRINT – IDLE (6 Single sided) ON ON
STANDBY SLEEP TO ON IDLE – PRINT – IDLE (3 double-sided) ON-app ON-app
OFF ON TO SLEEP IDLE – SCAN – IDLE (3 page) ON to SLEEP ON to SLEEP
IDLE – COPY – IDLE (6 single-sided) ON-app to SLEEP ON-app to SLEEP
IDLE – COPY – IDLE (3 double-sided) SLEEP SLEEP
SLEEP - IDLE SLEEP-app SLEEP-app
IDLE SLEEP to ON SLEEP to ON
SLEEP SLEEP-app to ON SLEEP-app to ON
ON to HIBERNATE ON to HIBERNATE
ON-app to HIBERNATE ON-app to HIBERNATE
HIBERNATE HIBERNATE
HIBERNATE-app HIBERNATE-app
HIBERNATE to ON HIBERNATE to ON
HIBERNATE-app to ON HIBERNATE-app to ON
ON to OFF ON to OFF
ON-app to OFF ON-app to OFF
CHARGING9 © 2014 Sekhar
Kondepudi
Field Deployment ( Faculty Offices)
• Desktops
• Laptops
• Monitors
• Printers
• Scanners
• MFDs
• VoIP Phones
10 © 2014 Sekhar Kondepudi
Some Average Measurements
11 © 2014 Sekhar Kondepudi
Desktops Transient Power Behavior (From Hibernate, Sleep, Off ) to ON
0 10 20 30 40 50 60 70 80 90 100 110 1200.00
20.0040.0060.0080.00
100.00120.00140.00160.00180.00200.00
Active Power vs TimeDesktop 1 Desktop 2 Desktop 3
Time (sec)
Activ
e Po
wer
(W)
SLEEP to ON
0 10 20 30 40 50 60 70 80 90 100 110 120 130 1400.00
20.0040.0060.0080.00
100.00120.00140.00160.00180.00200.00220.00240.00
Active Power vs TimeDesktop 1 Desktop 2 Desktop 3
Time (sec)
Activ
e Po
wer
(W
)
HIBERNATE to ON
0 10 20 30 40 50 60 70 80 90 100 110 1200.00
20.0040.0060.0080.00
100.00120.00140.00160.00180.00200.00220.00
Active Power vs TimeDesktop 1 Desktop 2 Desktop 3
Time (sec)
Activ
e Po
wer
(W)
OFF to ON
12 © 2014 Sekhar Kondepudi
Desktop Transient Current Draw
0 20 40 60 80 100 1200.000.100.200.300.400.500.600.700.800.90
Current vs TimeDesktop 1 Desktop 2 Desktop 3
Time (sec)
Curr
ent (
A)
SLEEP to ON
0 50 100 150 200 2500.000.050.100.150.200.250.300.350.400.450.50
Current vs TimeDesktop 1 Desktop 2 Desktop 3
Time (sec)
Curr
ent (
A)
HIBERNATE to ON
0 10 20 30 40 50 60 70 80 90 100 110 1200.00
0.10
0.20
0.30
0.40
0.50
0.60
Current vs TimeDesktop 1 Desktop 2 Desktop 3
Time
Curr
ent
OFF to ON
13© 2014 Sekhar Kondepudi
MFD Printers (Desktop, Monochrome)
0.00 20.00 40.00 60.00 80.00 100.00 120.00 140.00 160.000.00
100.00200.00300.00400.00500.00600.00700.00
Printer (MFD) - Idle State
Printer 1 Printer 2
Time (sec)
Pow
er (W
att)
5000 5200 5400 5600 5800 6000 6200 6400 6600 6800 70000
10
20
30
40
50
60
70
Printer (MFD) - Idle State
Time (sec)
Aver
age
Pow
er (W
)
Spike every 40 seconds (keep the drum warm ?)Spike lasts for 7-8 secondsConsumes up to 500W during that small timeframeTrend similar with Other MFD printer
Snapshot of Field Data Captured at 15 sec intervals over multiple weeksRange between 50 and 60 W Energy Consumption Consistent with Detailed Per Second Data
Assume 50 W for 12 hours a day X 365 Days X $ 0.25 / Kwh ~ $ 55.00 / year / printer At least 3000 such printers on NUS Campus = $ 165,000 savings annually
14 © 2014 Sekhar Kondepudi
Good Practice : Use Laptops over Desktops
• Laptops consuming 25% less power than desktops in ON mode and 50% less power during SLEEP mode.
• Our measurements show heat produced by desktops is 4 times more than laptops. Each laptop needs 3000 btuh less than a desktop to cool office space.
• Recommendation to procure Laptops when refresh time comes for computers
15© 2014 Sekhar Kondepudi
Popular Approaches for the Power Management of Desktops
• Wake-on-LAN : send packets on network to make machines sleep on wake up depending on network activity.
• SleepServer : transition to low-power sleep while maintaining network presence of all connected machines by a proxy mechanism on one server.
• LiteGreen : virtualize the desktop environment, migrating it between the user’s physical machine and virtual server.
16© 2014 Sekhar Kondepudi
Good Practice : Power Management of Desktops (1)
• Assigning power settings, more suited to the intended use of the computing devices.
• Many users of our building were not sure whether they should set their laptop in hibernate or sleep mode when they had to leave for lunch or home.
• When ON, consumption is 34 times more than in SLEEP and 116 times more than HIBERNATE.
• SLEEP consumes more power than HIBERNATE but enables a faster waking up time of the desktop. However the difference is quite tolerable • 20-65 sec, 0.46 Watts in Sleep and 40-80 sec, 0.08 Watts in Hibernate
17© 2014 Sekhar Kondepudi
Good Practice : Power Management of Desktops (2)Sleep vs. Hibernate
• Laptop Running NO applications. For less than 44 minutes and 30 seconds, it is better that a laptop is in SLEEP. For longer, hibernate is preferred.
Running applications. For less than 59 minutes and 39 seconds, it is better that a laptop is in SLEEP. For longer, hibernate is preferred.
• DesktopsThe tradeoff (independent of running applications or not) is 126 minutes and 3 seconds. Less than this time, it is better to put the desktop to SLEEP. More than this time, it is more practical to HIBERNATE.
Dilemma – Due to a large trade off of over two hours, the user is better off switching the desktop OFF when leaving it idle for over two hours.
18© 2014 Sekhar Kondepudi
SLEEP and OFF have nearly the same power consumption.
Hence switch off monitor, whenever a user puts their desktop in SLEEP.
Power consumed in IDLE is 15-22 times more than SLEEP.
Hence, set time to switch off display after 15 or so minutes of inactivity.
Good Practice: Power Management of Display
19© 2014 Sekhar Kondepudi
Good Practice: Brightness of Display
Adjusting brightness to 75% can lead to 18% decrease in consumption
Adjusting to 50%can lead to 30% savings.
Recommendation: 65%-85% brightness do not affect productivity while saving electricity.
20© 2014 Sekhar Kondepudi
Good Practice : Switch Printers OFF at night
• Common Printers & Multifunctional Devices (MFDs) in labs as well as those in the personal spaces of academic staff, remain IDLE/ SLEEP at night time.
• Especially at Night, there is no need to have the printers/ MFDs ready for tasks so these devices should be powered off, after hours.
0.00 20.00 40.00 60.00 80.00 100.00 120.00 140.00 160.000.00
100.00200.00300.00400.00500.00600.00700.00
Printer (MFD) - Idle State
Printer 1 Printer 2
Time (sec)
Pow
er (W
att)
Between $ 50 and $ 75 savings per printer per year
21© 2014 Sekhar Kondepudi
Good Practice : Single vs. Double Sided Printing
• Is double sided printing is more efficient compared , to single sided (in addition to the savings of paper ?
• YES – Always better to print double sided
Single-sided printing consumes 2.13 times more power than double sided.
0.124 Kwh for 3 double-sided pages compared to 0.058 Kwh for 6 single-sided pages.
22© 2014 Sekhar Kondepudi
MFD Double sided Copying
Time: 54 sec, Power: 0.209kWh Single Sided Copying
Time: 34 sec, Power: 0.117 kWh
Single-sided copying consumes 1.8 times less power than double-sided copying.
Hence, with regards to electrical energy savings, it is better to copy single-sided.
But there is a trade-off here, as single-sided copying uses more paper.
Good Practice : Single vs. Double Sided Copying
23© 2014 Sekhar Kondepudi
Good Practice : Scan+Email vs. Copy
• Is it better to Scan + Email or make a Physical Copy ?
• YES – Always better to Scan + Email : Energy + Paper savings
Sample of 6 pages scanned vs. 3 double-sided copies.
Copying needs 9 times more energy and 3 times more time.
24© 2014 Sekhar Kondepudi
Projectors have highest power consumption compared to other ICT devices. The projectors we tested consumes 220-285 Watts
SLEEP mode: Power: 10-12 Watts. Time to start up from SLEEP: 29 sec
OFF mode: Power: 0 Watts. Time to start up from OFF: 56 sec
Hence it is preferred to switch project OFF, as the 27 seconds saved in time, are not much compared to the gain in energy savings.
Projector Power consumption is directly linked to the brightness. Therefore it is suggested that only the needed LUMENS specifications must be installed.
Good Practice : Switching OFF projectors
25© 2014 Sekhar Kondepudi
We compared energy efficiency of an Energy Star labeled laptop and a non-labeled Laptop in both IDLE and SLEEP modes.
Energy Star Labeled Laptop consumes 43% less power than a non-labeled Laptop in
IDLE mode. 14% less power than a non-labeled Laptop in
SLEEP mode.
Hence, we strongly suggest to purchase Energy Star labeled ICT devices
Good Practice : Energy Star Equipment
26© 2014 Sekhar Kondepudi
Power consumption of VoIP phones is very low around 2.5 Watts for a VoIP.
Incoming or outgoing calls do not have an impact on the consumption of power.
1 9 17 25 33 41 49 57 65 73 81 89 97 105 113 121 129 137 1450
0.5
1
1.5
2
2.5
3
3.5
VoIP IDLE
VoIP CALL OUT
VoIP CALL IN
Time (seconds)
Po
we
r (W
att
s)
Even though VoIP usage is low, the phones consume power at night time when they are mostly idle.
Cisco has introduced an Energy-Wise feature, having the ability to turn the VoIP phone off and on based on the loads on the local network.
Good Practice : Energy Wise-Featured VoIP Phones
27© 2014 Sekhar Kondepudi
Good Practices particularly for University Campuses
28© 2014 Sekhar Kondepudi
An underestimated energy savings parameter is efficient and accurate utilization of common PC labs.
In SDE, during class in PC Lab time, the PC utilization is more aggressive, reaching 68%. This is still less to warrant a fully provisioned lab.
9:00 9:30 10:0010:3011:0011:3012:0012:3013:0013:3014:0014:3015:0015:3016:0016:3017:0017:3018:0018:3019:000
10
20
30
40
50
60
70
80Tuesday Thursday
Friday
Day Time
Oc
cu
pa
nc
y (
%)
Good Practice : Better Utilization of Common PC Labs
29
Class timings for the chart below:Tuesday 1100-1300, Thursday 1000 -1400
© 2014 Sekhar Kondepudi
Need to have Better Utilization of Common PC Labs
• Students use different machines each time.
• Students forget to switch them off after each use.
• 22% of the machines remain idle after use.
• 51% of the machines remain idle after use after a class. ON SLEEP OFF
0
10
20
30
40
50
60
70
Class No Class
Desktop State
Pe
rce
nta
ge
(%
)
Good Practice : Better Utilization of Common PC Labs
30© 2014 Sekhar Kondepudi
Educate students about the importance of saving energy by switching off the machines after use.
A Supervisor from students or staff, should make sure machines are switched off.
Differentiate general computing labs with PCs to specialized computing lab with PCs.
Specialized labs should only house as many students as course registration.
Generalized lab may offer just basic computing facilities to a larger student body.
Good Practice : Better Utilization of Common PC Labs
31© 2014 Sekhar Kondepudi
Some desktops in common pc labs, run simulations for hours or even a few days.
Friday
Thursday
Wednesday
Tuesday
Monday
0 5 10 15 20 25 30
SIMULATION ON
SLEEP OFF
NIGHT
Consumption (kWh)
Desktops used for simulations consume largest percentage of electricity. More than 50% on all weekdays.
Increased consumption at night time, due to desktops running simulations. Left ON from previous day.
Good Practice : Select desktops used in simulations wisely
32© 2014 Sekhar Kondepudi
Estimations of Potential Savings
• Almost $ 50K per year in School of Design & Environment
• University Wide – probably can extrapolate to $ 500 – 750 K per year
33© 2014 Sekhar Kondepudi
In Summary
• Great Potential to implement Best Practices which not only save energy but also money. Win-Win
• Areas of future workAudit power consumption of ICT infrastructural devices
Increase the duration of the study for greater than 6 months to identify temporal patterns for saving.
Associate user groups/profiles with use of ICT devices and try to develop more effective and personalized strategies to encourage occupants to adopt greener use of office equipment.
• Still not clear whether to involve building users more for energy savings or use automation. We support at least some education and training which is essential for users to perceive how to use their ICT devices.
34© 2014 Sekhar Kondepudi
Discussion
© 2014 Sekhar Kondepudi
35
Prof. Sekhar Kondepudi, [email protected]+65 9856 6472
Dr. Andreas [email protected]
The authors would like to acknowledge the support of the Ministry of Education, Singapore. Via an AcRF Grant for this project