Understanding the energy demands on mobile handsets - Embedded.com

Understanding the energy demands on mobile handsets

Simultaneous use of the diverse hardware systems embedded in a modern smart phone would limit many handsets to just a few hours of operation. In practice a phone will attempt to mitigate this prob- lem and extend its lifetime by making selective use of the available resources.

This is most often implemented through the use of standby power states, automatic control of the screen backlight, and actively switching particular subsystems (such as networking technologies) on or off as demand dictates.

These techniques are demand driven and so it is quite possible for a power-hungry application to drastically shorten the operating time of the handset.

Power-aware operating systems such as Cinder attempt to al- leviate this problem by enforcing energy allocations made to par- ticular processes. However, the complex and rapidly evolving way in which we interact with our handsets makes this allocation a difficult and dynamic problem.

The Google Nexus handset is a pertinant example. This device contains a 1GHz ARM CPU with additional hardware sup- port for various network technologies (e.g. GSM, UMTS, HSDPA, HSUPA, Wifi and Bluetooth), embedded GPS and A/V accelleration. Not only does this comprise a complex platform but there are often many different opportunities for achieving some particular goal each of which provides a different tradeoff in power consumption and performance.

This makes previous energy models and resources managers designed for laptops and data centers inapplicable. Applications such as Google Latitude create further complexity by generating correlated demand across many disparate subsystems of the phone.

In order to better understand the resource management chal- lenges posed by these devices we ran a preliminary study collect- ing data on handset usage from a small set of volunteers.

In this paper, we use our study to argue that system workload, resource utilisation and energy demands are diverse and dynamic both in time and space, are highly affected by contextual information, and vary significantly for individual users’ patterns of usage.

The ramifications of this are that the largely static, uncorrelated allocation systems used in systems such as ECOSystem and Cinder are likely to be very difficult to use in practice.

We highlight strict usage routines evidenced by some users as they interact ith their handsets at specific places and times. For these users it is possible that this kind of contextual information will prove a useful input to any energy allocation algorithm.

Our particular interest is in the construction of a Social Operating System which not only uses the hardware within a device to efficiently achieve some goal but which also shares this functionality between devices.

This study is our first work towards discovering the plausibility of such a system which will depend on the manner in which smart phone handsets are used, the demands of their applications and the energy costs thereof.

In this paper, we demonstrate the necessity of considering all those dynam- ics in order to characterise the energy demands of the system ac- curately.

These results indicate that simple algorithmic and rule-based scheduling techniques are not the most appropriate way of managing the resources since their usage can be affected by contextual factors, making necessary to find customised solutions that consider each user’s behaviour and handset features.

To read this external content in full, download the complete paper from the open online archives at the Technical University of Berlin.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.