Uploaded image for project: 'Observability and Data Analysis Program'
  1. Observability and Data Analysis Program
  2. OBSDA-1319

ARM Per-Process Power Metric Support

XMLWordPrintable

    • Icon: Feature Feature
    • Resolution: Unresolved
    • Icon: Undefined Undefined
    • None
    • None
    • PM Power-monitoring
    • None
    • False
    • Hide

      None

      Show
      None
    • False
    • Not Selected
    • 0

      Feature Overview (mandatory - Complete while in New status)
      Extend Power Monitoring (Kepler) to support ARM architecture machines utilizing the Hwmon (hardware monitoring) subsystem. Currently, Kepler relies solely on RAPL, which is not available on ARM platforms. ** This feature should enable process-level energy metrics for platforms supporting Hwmon similar to whats available for x86 environments today.
       
      Goals (mandatory - Complete while in New status)
      Implement Hwmon support as an alternative power meter source, enabling power monitoring on ARM platforms.
      What is the difference between today’s current state and a world with this Feature? Current State: There is no support for ARM plattforms
      Future State: Kepler can use data from ** Hwmon when RAPL is unavailable, providing power metrics on compatible ARM machines.
       
      Requirements (mandatory - _Complete while in Refinement status):

      Requirement Notes isMVP?
      Hwmon Integration: Implement logic to detect and read power metrics from the Hwmon subsystem. This must function as a fallback when RAPL/powercap is not present. Yes
      Metric Granularity: Hwmon data must be attributed to the Node, Pod, Container, and Process level. Must match the functionality of existing CPU process metrics. Yes
      Configurable Activation: The Hwmon functionality should be configurable   Yes

       
      Done - Acceptance Criteria (mandatory - Complete while in Refinement status): # Kepler successfully measures and attributes CPU power metrics (e.g., using zones like 'cpu power') on a verified ARM platform.

      1. Metrics are exposed at the process level and visible via the OpenShift Observe Console.

       
      Out of Scope (Initial completion while in Refinement status):_ * Resolving challenges related to *Heterogeneous Computing (machines with multiple core types) in terms of per-core power tracking.

      • Guaranteeing support or metric existence across all ARM vendor machines (e.g., Qualcomm, Altra) due to the variance in power metric availability.
      • Investigating further alternative power meters alongside RAPL and Hwmon.

              rh-ee-sherlofs Simon Herlofsson
              rh-ee-sherlofs Simon Herlofsson
              Votes:
              0 Vote for this issue
              Watchers:
              1 Start watching this issue

                Created:
                Updated: