Difference between revisions of "CELF Project Proposal/Add behavior-oriented power management"
(→Comments: add comment) |
(→Comments) |
||
| Line 62: | Line 62: | ||
unfortunately the firmware is not open. | unfortunately the firmware is not open. | ||
---- | ---- | ||
| − | Alan | + | Alan Carvalho de Assis wrote: |
I found this one: http://web.media.mit.edu/~msung/sak2.php | I found this one: http://web.media.mit.edu/~msung/sak2.php | ||
Revision as of 00:30, 15 December 2009
- Summary
- Add behavior-oriented power management
- Proposer
- Evgeny Ivashko
Contents |
Description
Every user has behavioral patterns. Behavioral pattern is the sequence of actions, running applications and so on. Some of these patterns are common and the others are individual. Knowing the patterns the program (that should be developed as the result of the project) can optimize the OS's behaviour to reduce energy consumption, response time and etc.
For example (common pattern), if user starts a video in the full-screen mode the program can automatically suspend some subsystems (like USB or CD/DVD-rom). Also the program can "hibernate" unused applications. Another example (individual pattern): if the user usually runs OpenOffice.org after the game or two of Mahjong, the program can automatically prepare the OS to quick start of the OOo.
Base
The work should be based on such technologies as Data Mining and Behavior Analysis
Project's objectives
The main result of the project should be the program (daemon) that can in (semi)automatic way change the power and application profiles, that should get
- reducing the device's energy consumption;
- reducing the OS's response time;
- reducing amount of used OS's resources.
Related work
This project should actively use results of such projects as:
- PowerTOP (http://www.lesswatts.org/projects/powertop/)
- BLTK (http://www.lesswatts.org/projects/bltk/)
The approach is well-proven in many different domains (http://en.wikipedia.org/wiki/Data_Mining#Notable_uses).
- Scope
- This should take about 6 months.
Comments
Patrick Bellasi writes:
That's a really interesting project. Moreover, I think it should be supported by an OpenSource HW reference platform which is cheap but sufficiently powerful to replace a semi-professional DAQ. ... I think that the signal acquisition layer should also support the "USB test and measurement" (TMC) class driver, which is already supported by Linux: drivers/usb/class/usbtmc.c
Patrick also wrote:
Probably the best one solution out there [for an OpenSource HW reference platform] I know is the USB-DUXfast, which is already supported by the COMEDI framework... but is missing the MTD usb class support. http://www.linux-usb-daq.co.uk/tech2_duxfast/ The board is relatively cheap compared to the offered features: up to 3MHz continous rate and up to 30MHz in burst mode (256samples), unfortunately the firmware is not open.
Alan Carvalho de Assis wrote:
I found this one: http://web.media.mit.edu/~msung/sak2.php
---