docs: update power attribution guide for v0.10+ rewrite#204
Open
iacker wants to merge 1 commit into
Open
Conversation
Addresses the gaps listed in issue sustainable-computing-io#199: - RAPL zones are described as discovered dynamically at runtime via the sysfs powercap interface, not from a hardcoded list. - Add a Platform Power Sources section covering RAPL, HWMon, and Redfish BMC (PowerSubsystem with fallback to the deprecated Power API) as node power sources. - Add a GPU Power Attribution section documenting NVML-based device power, idle/active split, per-process SM-utilization attribution, and MIG handling via DCGM with an equal-distribution fallback. - Note that the trained kepler-model-server does not yet integrate with the rewrite, which reads power directly from hardware interfaces. - Add GPU power metrics to Key Metrics and correct the outdated claim that GPU power is not captured. Fixes sustainable-computing-io#199 Signed-off-by: Billard <82095453+iacker@users.noreply.github.com>
✅ Deploy Preview for sustainable-computing ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Updates the power attribution guide to reflect the v0.10+ re-architecture, addressing the gaps listed in #199.
Changes
/sys/class/powercap/intel-rapl:*) instead of a hardcoded list, and notes primary-zone selection + name filtering.PowerSubsystemwith fallback to the deprecatedPowerAPI) as node power sources.nvmlDeviceGetPowerUsage), the idle/active split, per-process SM-utilization attribution (nvmlDeviceGetProcessUtilization), and MIG handling via DCGM with an equal-distribution fallback.kepler-model-serverdoes not yet integrate with the rewrite, which reads power directly from hardware.kepler_node_gpu_watts/kepler_process_gpu_wattsand corrects the outdated "GPU power not captured" statement.Notes
Documentation-only. Content was cross-checked against the current core implementation (
internal/device/RAPL/HWMon meters,internal/platform/redfish/, andinternal/device/gpu/nvidia/collector). I recently synced the developer-facing GPU design doc in the core repo (kepler#2498), so this keeps the user-facing guide consistent with it.Fixes #199