Skip to content

Releases: green-coding-solutions/green-metrics-tool

v2.1 AI/ML measurement features, Important Bugfix, Phase-Cut off mitigations

04 Jun 13:15
@ArneTR ArneTR
v2.1
741246d
This commit was signed with the committer’s verified signature.
ArneTR Arne Tarara
GPG key ID: 2540198A4079785B
Verified
Learn about vigilant mode.
Compare
Choose a tag to compare
v2.1 AI/ML measurement features, Important Bugfix, Phase-Cut off mitigations Latest
Latest

The new feature drops for Q2 in 2025 are coming to GMT. We know you could not wait any second longer :)

  • AI/ML Models are now a first class citizen in GMT for measurement. Check out our easy AI prompt measurement
  • Important Bugfix in the frontend where some values where shown as 0.0 See PR
  • Phase cut-offs with sampling are now easier to mitigate by pushing the phase end into the next sample
    • Too complicated? Read up on it in this PR - There are pictures, we promise!
  • AI Measurements and NVIDIA power is now done via our new C provider for less overhead

What's Changed

Full Changelog: v2.0...v2.1

Contributors

ribalba, ArneTR, and dependabot
Assets 2
Loading

v2.0 Variables in usage_scenarios, API changes and developer QoL

06 May 11:48
@ArneTR ArneTR
v2.0
fed833a
This commit was signed with the committer’s verified signature.
ArneTR Arne Tarara
GPG key ID: 2540198A4079785B
Verified
Learn about vigilant mode.
Compare
Choose a tag to compare
v2.0 Variables in usage_scenarios, API changes and developer QoL

The Green Metrics Tool has reached version 2.0! 🎉

This is because we have a major new functionality and also some breaking changes to API endpoints.

New functionalities:

  • Variables in Usage Scenario. You can use variables like __GMT_VAR_XXX__ in your usage_scenario.yml and supply them when you start a run. Either via CLI or handed in to the cluster. The scenario will then run with the respective replaced variable.
    Since comparing scenarios is a key feature of the Green Metrics Tool the usage_scenario variables are first class citizens and are a compare category. In short: same usage scenarios but different variables will create a separate category for comparison
    • Adding these also via the Web Frontend will come in v2.1
  • Badges have now selectable units between joules and watt-hours
  • Return data for watt-hours moved to milliwatt-hours
  • Added note column to be sent to Eco ci in case an error happens in the CI/CD pipeline
  • Eco CI now has a carbon badge in the repo overview. See example:
Screenshot 2025-05-06 at 1 47 53 PM - And man developer QoL fixes ... see the full changelog below

Breaking changes:

  • Badge endpoint does not accept general parameters like 'AC' anymore, but the concrete metric must be supplied. e.g. psu_carbon_ac_mcp_machine
  • runs endpoint migrated to v2 due to different return
  • run endpoint migrated to v2 due to different return
  • Usage scenario setup-commands key is now a list of dicts. See https://docs.green-coding.io/docs/measuring/usage-scenario/
  • Ignore unsupported docker compose keys. Can be turned off with ignore-unsupported-compose: true

What's Changed

Read more

Contributors

ribalba, ArneTR, and 2 other contributors
Loading

v1.4.0 - WattHours default + QoL + depedency updates

10 Feb 08:03
@ArneTR ArneTR
v1.4.0
d923b92
This commit was signed with the committer’s verified signature.
ArneTR Arne Tarara
GPG key ID: 2540198A4079785B
Verified
Learn about vigilant mode.
Compare
Choose a tag to compare
v1.4.0 - WattHours default + QoL + depedency updates

If you look at the release logs for this month you might be thinking that only dependency updates have happened. But don't be fooled, a lot of long wanted and interesting features have made it to this months release:

  • WattHours are now the default display (instead of Joules) as the unit of energy. You can still switch back to Joules in the settings
  • QoL features in the dashboard:
    • Unselect button when comparing runs
    • href removed for non links for screen readers
    • Badges now have average function
  • GMT can now isolate containers from network by setting --internal
  • GMT is now microjoules under the hood to allow higher resolutions
  • Averages are now weighted averages for CPU utilization and others
  • Sample resolution is now shown in Dashboard and you can compare if you can actually sample at the target rate

What's Changed

Full Changelog: v1.3.0...v1.4.0

Contributors

ribalba, ArneTR, and dependabot
Loading

v1.3.0 Codespaces + CarbonDB 2.0 + QoL / Tests / Package updates

17 Dec 19:35
@ArneTR ArneTR
v1.3.0
fb8a46a
This commit was signed with the committer’s verified signature.
ArneTR Arne Tarara
GPG key ID: 2540198A4079785B
Verified
Learn about vigilant mode.
Compare
Choose a tag to compare
v1.3.0 Codespaces + CarbonDB 2.0 + QoL / Tests / Package updates

Whoo, what a christmas release!!

GMT comes along for the holidays with an unusual packed feature release and also some more extensive test coverage now even including the frontend.

The higlights are:

  • We now have a default Codespaces setup to try at your convenience. Just click on the Codespaces button in the Readme and try out GMT pre-installed and pre-configured with our trusty Cloud Energy reporter active for cloud measurements
  • The integrated Cloud Energy now accounts vor vhost ratio of your machine so you can get reliable GMT values even in VMs
  • Phase Stats are now also saved if runs fails. So no data gets lost if you want to account for the carbon consumption of your measurement cluster
  • CarbonDB 2.0 is now ready for use. (Please note that this is an enterprise only feature and distributed in a separate repository - See https://www.green-coding.io/products/carbondb/)
  • Clickable Comparsion Charts + Display Missing Values by @ArneTR in #1017 So nice!!!!!

Thanks again to our contributors this month:

What's Changed

Full Changelog: v1.2.0...v1.3.0

Contributors

ribalba, ArneTR, and 2 other contributors
Loading

v1.2.0 STDDEV changes and display tweaks

14 Oct 13:14
@ArneTR ArneTR
v1.2.0
d098fa1
This commit was signed with the committer’s verified signature.
ArneTR Arne Tarara
GPG key ID: 2540198A4079785B
Verified
Learn about vigilant mode.
Compare
Choose a tag to compare
v1.2.0 STDDEV changes and display tweaks

STDDEV in GMT is now exclusively calculated as SAMPLE-STDDEV and not POP-STDDEV anymore.

This is for two reasons:

  • It is required by certifications like the Blue Angel for Software
  • We got many debates that in cases where the average is only estimated through measurements and is not absolute one MUST use the sample STDDEV. Still one could argue that one does not want to characterize the measured software but rather the measurement setup it is safer to use the sample STDDEV as it is always higher

Furthermore some display changed where made to have the AVG bar optional and quick-access for the timeline was added on overview

What's Changed

Full Changelog: v1.1.1...v1.1.2

Contributors

ArneTR and dependabot
Loading

v1.1.1 SCI / Embodied Carbon Patch & dependencies update

07 Oct 12:23
@ArneTR ArneTR
v1.1.1
de56353
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.
Compare
Choose a tag to compare
v1.1.1 SCI / Embodied Carbon Patch & dependencies update
  • (fix) SCI values were not correctly reported anymore.
  • (update) Using the release to also update some depedencies

What's Changed

Full Changelog: v1.1...v1.1.1

Contributors

ArneTR and dependabot
Loading

v.1.1.0 - Patch Release / Feature Release

05 Oct 08:48
@ArneTR ArneTR
v1.1
2ceb2cb
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.
Compare
Choose a tag to compare
v.1.1.0 - Patch Release / Feature Release

v1.0 is just a week out, and here it is: The first patch / feature release :)

We encourage everybody to update as one reporter was reporting broken data.

Fixes

  • Some API calls in the HOG are now optional as we have seen machines without disks
  • GMT uses for reporiting the SI norms. That means that 1 KB is 1000 Bytes, not 1024 (as this would be kiB). We fixed one occurence where this was not correctly reported but should be negligeable to most users.
  • network_io_cgroup_container reporter was reporting grossly false high values. Here an accumulation instead of an assign was happening. (Note that this was introduced in v1.0 and was fine in versions before. Blast radius should thus be small)

Features addons

  • SSL is now possible in the installation process. It should be pretty straight forward but documentation is to follow.
  • Cron dir is now refactored and all cron jobs have error handling
  • Added a backup script to import backups to the database (Can handle 100 GB+ imports !)

Important info

CarbonDB 2.0 is actually not part of this release. Due to some branch merging it looks this way on GitHub. It will be released later next week with v1.2

What's Changed

Full Changelog: v1.0...v1.0.1

Contributors

ribalba, ArneTR, and dependabot
Loading

v1.0

30 Sep 14:26
@ArneTR ArneTR
v1.0
332347f
This commit was signed with the committer’s verified signature.
ArneTR Arne Tarara
GPG key ID: 2540198A4079785B
Verified
Learn about vigilant mode.
Compare
Choose a tag to compare
v1.0

The 1.0 release !!!! 🎉

While this release mostly contains updates and tests to asset the stability of the measurements it has two main perks:

  • GMT now has an authentication so it can be used with multiple users to implement ACLs and data retention
  • GMT is now fully compliant with the measurement requirements for the new Blue Angel for Software

We opted for the 1.0 release now as we have used it to capture more than 1 TB of measurement data of 100+ users.

GMT has been featured on many conferences and papers

We believe it is now ready for 1.0. Enjoy!

What's Changed

Full Changelog: v0.25...v1.0

Contributors

ribalba, ArneTR, and 2 other contributors
Loading

v0.25 Security Release! + Feature drop

25 Jul 11:03
@ArneTR ArneTR
v0.25
a328104
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.
Compare
Choose a tag to compare
v0.25 Security Release! + Feature drop

Security

When install GMT in an unsecured multi-user environment a malicious actor that has already penetrated the system could replace a binary in an allowed sudoers path to get root access rights.

This security vulnerability is to be seen as relatively low, as the system is not open from the outside. It is a local privilege escalation.

We recommend updating to this release as soon as possible. Fix can be applied by just running the install script (install_linux.sh / install_mac.sh) again.

Features

  • Power per Container is now displayed in the dashboard. This feature is based on a linear estimation of the container CPU usage and the total machine power
  • Multi-Stage-Build Docker Images Support
  • Global Timeout to abort measurement when a timer is expired
  • Weekly projects for the timeline projects to run a project only once a week
  • RAPL energy filtering guard clause is now checked

What's Changed

Read more

Contributors

ribalba, ArneTR, and 2 other contributors
Loading

v0.24 - Optimizations, Diffing, Intel PSYS, Performance, CarbonDB & QoL

22 Apr 10:17
@ArneTR ArneTR
v0.24
0187979
This commit was signed with the committer’s verified signature.
ArneTR Arne Tarara
GPG key ID: 2540198A4079785B
Verified
Learn about vigilant mode.
Compare
Choose a tag to compare
v0.24 - Optimizations, Diffing, Intel PSYS, Performance, CarbonDB & QoL

Oh my, what a release!

The most important: Green Metrics Tool now has optimizations!

When you do a run you will be informed about what to tweak and optimize in order to make your software greener. We are releasing a first bunch of battle tested optimizations into open source. More to follow!

See details on the bottom of the GMT page here: https://www.green-coding.io/projects/green-metrics-tool/

Other very prominent changes are:

  • Introducing heavy caching with Redis for API Requests, badges and internal calculations
  • Error handling is now modular and will use a Queuing system if wanted. Blocking email errors are gone!
  • CarbonDB added. You can now find out how much carbon your total cluster is using!
  • Diffing has been implemented! See details here: #693
  • PSYS domain can now be captured to get the whole energy of your Notebook: https://docs.green-coding.io/docs/measuring/metric-providers/psu-energy-dc-rapl-msr-machine/
  • Many checks for measurement stability, taskset in order to reduce overhead, docker builds have now layer caching for faster GMT updates, DB reconnects and much more. Read the detailes under What's changed

What's Changed

Full Changelog: v0.23...v0.24

Contributors

ribalba and ArneTR
Loading