Power attribution and Estimation Engine in Linux
Worked on development of the power panel in GNOME-Usage with Felipe Borges and Christian Kellner.
Project Log
- Problem and Motivation: Why: not all software is created equal. Some are more harmful than others, just like website have different footprints. We want fairness in energy consumption via quantification for each process
- The What, Who, Why, When, Where, & How of the Power Measurement problem
- Windows Energy Estimation Engine breakdown: Link
- Windows System Power Report : Video 1, Video 2
- MAC OS X Activity Monitor Power Impact breakdown: Link
- Android Power Measurement system : Unknown!
- Reverse Engineering PowerTOP : Link
- Understanding Intel RAPL: Link
- Developing the power models, collecting and crunching global user data
- UPower - System Architecture and Integrations
- Linux Process accounting infrastructure and integrating power event recording
- turbostat - Accessing CPU and GPU Package details and power: Link
- Comparison and embedding into sysprof –> Right now gives only CPU and GPU usage, whereas power is a multi-faceted problem
- Update SuperUser question for all power models: Link
- Next step = GPU in GNOME-Usage. GSoC Proposal
- Intel CPU Energy analysis tool
Known tools (from LFE Summit)
Kernel Developer
Relevant Research Papers
- Performance is NOT same as energy – https://www.sciencedirect.com/science/article/pii/S0306261920304694
- Good paper on power impact of managed languages: https://users.cecs.anu.edu.au/~steveb/pubs/papers/powerperf-asplos-2011.pdf
- https://www.microsoft.com/en-us/research/publication/empowering-developers-to-estimate-app-energy-consumption/
- AppScope: Application Energy Metering Framework for Android Smartphone Using Kernel Activity Monitoring
- DRAM Power Measurement with RAPL
- RAPL for memory
- https://ieeexplore.ieee.org/document/5751489
- http://www.cs.columbia.edu/~lierranli/coms6998-11Fall2012/papers/eprof_eurosys2012.pdf
- https://dl.acm.org/doi/pdf/10.1145/3194078.3194079
- https://dl.acm.org/doi/10.1145/3229631.3229633
- https://dl.acm.org/doi/pdf/10.5555/2872599.2872613
- https://dl.acm.org/doi/10.1145/3575693.3575709
- https://dl.acm.org/doi/10.1145/3575693.3575754
Data collection sources
- Tracy: Can collect data from HPC systems for workloads to fine-tune the model
Resources
- Linux packages to poll information: https://manpages.debian.org/testing/linux-cpupower/index.html
- https://apenwarr.ca/log/20061024
- Linux ACPI Intel Documentation: Link
- Linux process management: Link
- Linux Process Model (from 2000) : Link
- Smart Battery Systems: Link
- Hacking Battery Firmware for fun and profit: Link
- https://www.semanticscholar.org/paper/Adaptive-Power-Management-Based-on-Reinforcement-Liu-Hsu/40b943ed1f4514979b2b997c3a675cb77ec635c7
- Is it possible to compute carbon footprint of the current application: https://css-tricks.com/reduce-your-websites-environmental-impact-with-a-carbon-budget/ ?
- https://www.phoronix.com/scan.php?page=news_item&px=GNOME-Usage-Power-2020
- https://utcc.utoronto.ca/~cks/space/blog/linux/LoadAverageWhereFrom
- https://eclecticlight.co/2022/05/02/dont-trust-activity-monitor-on-m1-macs/
- https://medium.com/codex/what-are-the-greenest-programming-languages-e738774b1957
- https://sites.google.com/view/energy-efficiency-languages/home
Links from LFE
- https://greensoftware.foundation/projects/
- https://greensoftware.foundation/articles/software-carbon-intensity-sci-specification-project
- https://codecarbon.io/#howitwork
- https://news.microsoft.com/de-ch/2023/01/10/carbon-aware-computing-whitepaper/
- https://blog.google/inside-google/infrastructure/data-centers-work-harder-sun-shines-wind-blows/
- Toward Carbon-Aware Networking, Noa Zilberman, University of Oxford
- https://news.microsoft.com/de-ch/2023/01/10/carbon-aware-computing-whitepaper/
- https://www.microsoft.com/en-gb/industry/blog/technetuk/2021/10/12/how-to-measure-and-reduce-the-carbon-footprint-of-your-application/
