We distinguish power from energy for clarity. Power is the
rate of energy consumption, i.e. energy = power × time.
For the metrics, we use watts (W) for power, and watt-hour
(Wh) instead of joule (J) for energy. As shown in Fig. 2, we
used six typical commodity desktops and laptops for clients
whose specifications and idle power usage are described in
Table 1. We have measured power usage on various types of
PCs from legacy (L1 and D1) and standard (L2 and D2) to
high-performance (L3 and D3). The measured power of PCs
in this paper includes the monitor for the fair comparison
with laptops. We considered MS Internet Explorer (IE) 8
for the main web browser because of its wide popularity.
Test computers are connected to the campus network with
100 Mbps Ethernet.
To measure the power usages of laptops or PC, as well as
monitors, we used the“Wattsup PRO .net [22]” device. The
Wattsup PRO .net box monitors power, current, and voltage
information of testing devices, and it writes the power in
W, or the energy in Wh, to its memory every two seconds.
This power information on the Wattsup PRO .net device
can be sent to a server for further analysis by USB or LAN.
For this purpose, we wrote a Python script program that
periodically retrieves the power consumption data from the
Wattsup PRO .net device.
For each test run collecting power consumption information
from a test PC, an automated script triggers a web
browser to visit the top page of a website for the given timeperiod (e.g., two minutes). We capture all the packets for
each website in order to examine the web contents. For each
website, we collect user statistics from a 3rd-party information
provider, Bizinformation. Before moving to a new website
for a test, we delete the web cache and put the browser
into the idle state for a short period (e.g., two minutes).
By deleting web caches and idling the PC for two minutes,
we can maintain the fair experimental environment for every
run of experiments. We repeated the website test 10
times and averaged the power consumption value (2011.2.20
- 2011.3.10). We have investigated popular top 100 websites
in China, Japan, Korea, and the US at the Alexa site [7].
Though the number of sample websites from these countries
is not enough to reflect the pattern of global websites, our
experiments with 400 websites can be an indicator for the
energy consumption of websites.
We distinguish power from energy for clarity. Power is therate of energy consumption, i.e. energy = power × time.For the metrics, we use watts (W) for power, and watt-hour(Wh) instead of joule (J) for energy. As shown in Fig. 2, weused six typical commodity desktops and laptops for clientswhose specifications and idle power usage are described inTable 1. We have measured power usage on various types ofPCs from legacy (L1 and D1) and standard (L2 and D2) tohigh-performance (L3 and D3). The measured power of PCsin this paper includes the monitor for the fair comparisonwith laptops. We considered MS Internet Explorer (IE) 8for the main web browser because of its wide popularity.Test computers are connected to the campus network with100 Mbps Ethernet.To measure the power usages of laptops or PC, as well asmonitors, we used the“Wattsup PRO .net [22]” device. TheWattsup PRO .net box monitors power, current, and voltageinformation of testing devices, and it writes the power inW, or the energy in Wh, to its memory every two seconds.This power information on the Wattsup PRO .net devicecan be sent to a server for further analysis by USB or LAN.For this purpose, we wrote a Python script program thatperiodically retrieves the power consumption data from theWattsup PRO .net device.For each test run collecting power consumption informationfrom a test PC, an automated script triggers a webbrowser to visit the top page of a website for the given timeperiod (e.g., two minutes). We capture all the packets foreach website in order to examine the web contents. For eachwebsite, we collect user statistics from a 3rd-party informationprovider, Bizinformation. Before moving to a new websitefor a test, we delete the web cache and put the browserinto the idle state for a short period (e.g., two minutes).By deleting web caches and idling the PC for two minutes,we can maintain the fair experimental environment for everyrun of experiments. We repeated the website test 10times and averaged the power consumption value (2011.2.20- 2011.3.10). We have investigated popular top 100 websitesin China, Japan, Korea, and the US at the Alexa site [7].Though the number of sample websites from these countriesis not enough to reflect the pattern of global websites, ourexperiments with 400 websites can be an indicator for theenergy consumption of websites.
การแปล กรุณารอสักครู่..
