Uploaded image for project: 'OpenShift Bugs'
  1. OpenShift Bugs
  2. OCPBUGS-44562

[OCP] CPU Ratio of the Z Architecture for OpenShift

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Unresolved
    • Icon: Critical Critical
    • None
    • 4.14, 4.15, 4.16, 4.17
    • Documentation
    • None
    • Moderate
    • None
    • False
    • Hide

      None

      Show
      None

      Description of problem: 

      Many Red Hat products require the same number of vCPUs of the x86 environment even if the Z's Integrated Facility for Linux (IFL) processor is much more powerful.For example, the Red Hat OpenShift Data Foundation requires the same vCPU for both Z and x86 as highlighted at the link below: 
      
      https://docs.redhat.com/en/documentation/red_hat_openshift_data_foundation/4.14/html/planning_your_deployment/infrastructure-requirements_rhodf#platform-requirements_rhodf
      
      This point can be also observed in the "Minimum resource requirements for cluster installation" of Red Hat OpenShift.
      
      # for IBM-Z
      https://docs.openshift.com/container-platform/4.17/installing/installing_ibm_z/installing-ibm-z.html#installation-minimum-resource-requirements_installing-ibm-z
      # for UPI over vSphere
      https://docs.openshift.com/container-platform/4.17/installing/installing_vsphere/upi/upi-vsphere-installation-reqs.html#installation-minimum-resource-requirements_upi-vsphere-installation-reqs 
      
      In the Customer real deployment with 3 ODF nodes with 12 vCPU, the real vCPU consumption was less then 1 vCPU for each ODF node. 
      
      Defining the required resources aligned to x86, this approach leads to a situation where several resources (vCPU) are wasted because underused and reserved to the OCP nodes.
      
      IBM literature (but also independent sources on the net) decree a consolidation ratio between 10 and 30 x86 cores for IFL on various workloads. 
      
      Below is an IBM document that publicly states this: https://www.ibm.com/downloads/cas/G3QK9D8V
      
      Also interesting is this other document: https://www.ibm.com/downloads/cas/X6OLNMGE
      On workloads like mongo DB you generally get to more than 1:20
      
      Based on these facts, a review of the OpenShift and operators requirements related to Z architecture-based environments is recommended (for example: a 1:3 ratio in the favor of IBM-Z) to prevent any waste of vCPU resources.
      

       

      Version-Release number of selected component (if applicable):

      Red Hat OpenShift 4.14 and next minor release

      How reproducible:

          OCP Deployment over IBM-Z with Integrated Facility for Linux (IFL) processor

      Steps to Reproduce:

          1.
          2.
          3.
          

      Actual results:

       vCPU wasted because undersized and reserved to the OCP nodes

      Expected results:

       Review of the OpenShift and operators requirements related to Z architecture-based environments

      Additional info:

          

       

              ocp-docs-bot OCP DocsBot
              rhn-support-rbruzzon Riccardo Bruzzone
              Votes:
              0 Vote for this issue
              Watchers:
              1 Start watching this issue

                Created:
                Updated: