Open Source Subnet An independent Open Source community View more

AT&T researchers call for smartphone apps that won't suck your battery dry

Different OSs may have different energy efficiency

Who doesn't love Pandora Radio? But listening to it on my Android phone is the fastest way to kill the battery and what good is a mobile phone if it has to be constantly plugged in? New research shows that Android phones are the most data hungry smartphones out there. A group of researchers at AT&T Labs are calling on app makers to fix this problem by building more energy-aware apps. Not surprisingly, Pandora is one of their test subjects. (Facebook is another).

It's about time! All smartphones could use more energy efficient apps, but Android users suffer the most. New research from Nielsen found that although iPhone users engage in as much or more data-intensive activities (downloading apps, streaming music or video), as Android users, Android phones gobble up more monthly data. Each month, Androids are consuming 90MB more data than iPhone users.

As every smartphone user knows, the more data transferred, the faster the battery drains. But apparently it's not the OS that's the issue ... it's the underlying apps, researchers say.

Enter new research being done by AT&T to create energy-efficient apps that recognize they are on a cell network, and limit both the number of times an app connects to the network and the time needed to connect. They have developed a tool that helps app developers figure out when their apps really need full power connections (download speeds of around 7.1Mbits/sec) or when the app can get by on a proposed "intermediate state" which consumes half the power and transmits less data at a slower speed, typically by sharing a low-speed channel, (often 16kbps).

For instance, the researchers found that when they ran Pandora for 12 minutes, the app conducted a series "of short bursts once every 62.5 seconds ... While the music itself was sent simply and efficiently as a single file, the periodic audience measurements — each constituting only 2KBs or so — were being transmitted at regular 62.5-second intervals. The constant cycle of ramping up to full power (2 seconds to ramp up, 1 second to download 2KB) and back to idle (17 seconds for the two tail times ... was extremely wasteful," they wrote.

After reading the paper, I had many questions. I contacted the researcher Alexandre Gerber, a principal member of the technical staff at AT&T Labs Research, and asked them.

How do the different platforms compare when it comes to energy efficiency already? Apple/iPhone/ios vs. Android vs. BlackBerry vs. Windows Phone 7. Does development platform influence this? Are some better than others?

Different OSs may have different energy efficiency in terms of some system components such as CPU and memory. But we are looking at the efficiency of accessing network (3G power consumption contributes about 50% of overall handset power consumption). That is mostly determined by the application instead of OS.

How do different network speeds/types influence this … 3G vs. 4G vs. WiFi?

It is the resource control policy of different networks that influence the energy efficiency. Cellular networks usually have similar resource control mechanisms, but cellular network technology is also getting better over time. WiFi has a different approach and it is more energy efficient than cellular.

The paper mentions, “One popular app was found to be using 40% of its power consumption to transmit 0.2% of its data.” … was this a typical finding? Or was it an extreme finding?

This is a common observation for applications with periodic data transfers (e.g. ads, keep alive, pull instead of the more efficient push, audience measurements), although the numbers may not always be that high.

In terms of hours of battery life, how much power overall would you guess is wasted by apps that do a poor job of managing state? (What I mean is, if you have a battery that is supposed to give you six hours of talk time, but dies in three hours of app usage, how much battery life would you get back if all of your apps were energy efficient? A few minutes? A few hours?)

Clearly that depends on the application you are using. For a large Internet radio, for instance, if 40% of its radio power, which contributes to 50% of total device power, is wasted, then you can save about 40% * 50% = 20% of overall battery life. So this could end up being a significant amount of time.

How much is app battery usage influenced/dependent on the handset? Do the same apps consume different amounts of energy on different handsets  (and HTC android phone vs. a Motorola one? An iPhone 3 vs an iPhone 4)?

Yes they differ.  The table below compares power consumption of three radio states (IDLE, FACH, and DCH) of two phones: HTC TyTn II and Google NexusOne. These are measurements made as part of our study in our Research group; it is independent of measurements made by our official device testing group:

Radio State | TyTn | NexusOne

-----------------------------------------

P(IDLE) | 0 | 0

P(FACH) | 460mW | 450mW

P(DCH) | 800mW | 600mW

In your paper, you detailed the results of analyzing the Pandora app, what smartphone platform did you use to analyze it?

Google Android.

Generally speaking, did you discover the Facebook app was more (or less) energy efficient than Pandora?

This is an apple to orange comparison. These are applications that are difficult to compare; the content is completely different. A comparison would only make sense between the same type of applications. For instance, we noticed that Pandora is more efficient than other Internet radios because it is sending data in bursts followed by long periods of inactivity, as opposed to continuously streaming content like some other Internet radios.

 

More from Open Source Subnet

Smartphone apps will boost enterprise productivity by 2015

Apache president Jim Jagielski talks about OpenOffice.org next steps

Does Amazon "owe" open source? Maybe a little

Fun with Linux: How to wipe out Windows to install Ubuntu, then decorate with a penguin

Top 12 Google Labs experiments

Follow all Open Source Subnet bloggers on Twitter @OSSubnet

Follow Julie Bort on Twitter @Julie188

 

Open Source Subnet

Join the discussion
Be the first to comment on this article. Our Commenting Policies