3G v.s WIFI Radio Energy with YouTube downloads

1
3G v.s WIFI Radio Energy with YouTube downloads
2
Energy in Mobile Phone Data Transfers

• In 3G, there are three states
• Idle
• DCH (Dedicated Channel), do not share the channel
  with any other device. Used for high data rate
  and low delay when downloading a large file.
• FACH (Forward Access Channel), share the channel
  with other devices. Use when only a low data rate
  is needed
• Power
• DCH uses the most power
• FACH is half of what DCH uses
• Idle uses about 10 of FACH

• State Transition
• Transition from idle to DCH takes about 4
  seconds. This is delay and wasted power.
• Transition from FACH to DCH does not have this
  wasted energy.
• One issue is how long of idle activity until
  switching tom DCH to FACH and from FACH to idle
• (Note that idle after the download is not the
  same as idle before download)
• WIFI
• WIFI requires much more energy to associate with
  an access point that the transition to DCH does.
• However, WIFI has an efficient power save mode,
  like the FACH, but much lower power. (I don't
  have numbers. Perhaps this can be a project)

3
YouTube via 3G and Wifi

• Youtube uses progressive download, not streaming
• Three phases
• Initial download file is downloaded but not
  played. Instead, cache is filled
• Download and play
• Play only
• Most likely, the 3G includes transitioning to
  DCH.
• This can be observed in the rmap up to full power
  It appears that the WIFI was already
  associated.

4
Playback from Local Storage

• Three options
• Playback from flash storage
• Playback from memory
• Playback from on-chip cache (replay)
• Note that some energy is needed to move data from
  memory to on-chip cache. Once the data is on the
  chip, the power usage is low.
• Playback vs download
• Since the radios are off for this playback, the
  energy usage is lower than the "play-only" phase
  of the progressive download.
• On the other hand, the energy usage is not that
  much lower, especially in the wifi case. This
  shows that in mobile phones, wifi radio on uses
  much less than the CPU, display, etc.
• Total energy from playback
• From memory 281.078J.
• From Flash 281.136J
• From cache (replay) 275.128J
• In all cases, about 1/3 of download form 3G, 1/2
  of download from wifi

5
Impact of Temperature

• Erratic variation in energy.
• However, there is little change in temperature.
• Perhaps other changes in temperature could cause
  larger impacts.
• Then again, how would temperature impact energy
  usage.
• It could impact battery capacity, but should not
  impact energy usage. So why the variation?

6
Download and Then Play

• There is little different between progressive
  download and two-phase
• The power during 3G play-back in the two-phase is
  less than the power used in play-back only phase.
  
• This is because the radio is in FACH state during
  much of the playback-only phase
• However, playing from local storage uses less
  energy than the second phase of the two-phase
  download. This does not make sense

7
Fun with 3G, WIFI and Bluetooth

• Wireless wakeups revisited energy management for
  voip over wi-fi smartphones, Agarwal et al,
  2007.
• Proposed an architecture to use the GSM radio to
  wake up the WiFi radio upon an incoming VoIP call
  to leverage the better quality and
  energy-efficiency ofWiFi while keeping its
  scanning costs low.
• Efficient and transparent dynamic content updates
  for mobile clients, Armstrong et al, 2006.
• Application level modifications to reduce energy
  consumption for updates to dynamic web content.
• Use a proxy to
• only push new content when the portion of the web
  document of interest to the user is updated,
• batch updates to avoid the overhead of repeated
  polling, and
• use SMS on GSM to signal the selection of WiFi or
  GSM based on the transfer size for
  energy-efficient data transfer.
• Reducing the power consumption of wireless mobile
  devices with multiple radio interfaces, Pering,
  2006.
• Strategies to intelligently switch between WiFi
  and Bluetooth.
• Context-for-wireless context-sensitive
  energy-efficient wireless data transfer, Rahmati
  et al. 2007.
• Show that intelligently switching between WiFi
  and GSM reduces energy consumption substantially
  as WiFi consumes less transmission power.
  However, in order to avoid the cost of
  unnecessary scanning in the face of poor WiFi
  availability, the authors design an algorithm
  that predicts WiFi availability, and the device
  scans forWiFi access points only in areas
  whereWiFi is available with high probability.
• Energy Consumption in Mobile Phones A
  Measurement Study and Implications for Network
  Applications, Balasubramanian et al, 2009,
• Schedule download to be in bulk, and hence save
  on the start-up and FACH energy. Note that these
  energies can be relatively large when small files
  are downloaded