This is a cache of https://docs.okd.io/4.15/scalability_and_performance/telco_ref_design_specs/ran/telco-ran-du-overview.html. It is a snapshot of the page at 2024-11-26T21:12:38.270+0000.
Telco RAN DU use model overview - Reference design specifications | Scalability and performance | OKD 4.15
  1. Documentation
×

Telco RAN DU use model overview

Use the following information to plan telco RAN DU workloads, cluster resources, and hardware specifications for the hub cluster and managed single-node OpenShift clusters.

Telco RAN DU application workloads

DU worker nodes must have 3rd Generation Xeon (Ice Lake) 2.20 GHz or better CPUs with firmware tuned for maximum performance.

5G RAN DU user applications and workloads should conform to the following best practices and application limits:

  • Develop cloud-native network functions (CNFs) that conform to the latest version of the CNF best practices guide.

  • Use SR-IOV for high performance networking.

  • Use exec probes sparingly and only when no other suitable options are available

    • Do not use exec probes if a CNF uses CPU pinning. Use other probe implementations, for example, httpGet or tcpSocket.

    • When you need to use exec probes, limit the exec probe frequency and quantity. The maximum number of exec probes must be kept below 10, and frequency must not be set to less than 10 seconds.

Startup probes require minimal resources during steady-state operation. The limitation on exec probes applies primarily to liveness and readiness probes.

Telco RAN DU representative reference application workload characteristics

The representative reference application workload has the following characteristics:

  • Has a maximum of 15 pods and 30 containers for the vRAN application including its management and control functions

  • Uses a maximum of 2 configmap and 4 Secret CRs per pod

  • Uses a maximum of 10 exec probes with a frequency of not less than 10 seconds

  • Incremental application load on the kube-apiserver is less than 10% of the cluster platform usage

    You can extract CPU load can from the platform metrics. For example:

    query=avg_over_time(pod:container_cpu_usage:sum{namespace="openshift-kube-apiserver"}[30m])

  • Application logs are not collected by the platform log collector

  • Aggregate traffic on the primary CNI is less than 1 MBps

Telco RAN DU worker node cluster resource utilization

The maximum number of running pods in the system, inclusive of application workloads and OKD pods, is 120.

Resource utilization

OKD resource utilization varies depending on many factors including application workload characteristics such as:

  • Pod count

  • Type and frequency of probes

  • Messaging rates on primary CNI or secondary CNI with kernel networking

  • API access rate

  • Logging rates

  • Storage IOPS

Cluster resource requirements are applicable under the following conditions:

  • The cluster is running the described representative application workload.

  • The cluster is managed with the constraints described in "Telco RAN DU worker node cluster resource utilization".

  • Components noted as optional in the RAN DU use model configuration are not applied.

You will need to do additional analysis to determine the impact on resource utilization and ability to meet KPI targets for configurations outside the scope of the Telco RAN DU reference design. You might have to allocate additional resources in the cluster depending on your requirements.
Additional resources

Hub cluster management characteristics

Red Hat Advanced Cluster Management (RHACM) is the recommended cluster management solution. Configure it to the following limits on the hub cluster:

  • Configure a maximum of 5 RHACM policies with a compliant evaluation interval of at least 10 minutes.

  • Use a maximum of 10 managed cluster templates in policies. Where possible, use hub-side templating.

  • Disable all RHACM add-ons except for the policy-controller and observability-controller add-ons. Set Observability to the default configuration.

    Configuring optional components or enabling additional features will result in additional resource usage and can reduce overall system performance.

    For more information, see Reference design deployment components.
Table 1. OpenShift platform resource utilization under reference application load
Metric Limit Notes

CPU usage

Less than 4000 mc – 2 cores (4 hyperthreads)

Platform CPU is pinned to reserved cores, including both hyperthreads in each reserved core. The system is engineered to use 3 CPUs (3000mc) at steady-state to allow for periodic system tasks and spikes.

Memory used

Less than 16G

Telco RAN DU RDS components

The following sections describe the various OKD components and configurations that you use to configure and deploy clusters to run telco RAN DU workloads.

A diagram describing the telco RAN DU component stack.
Figure 1. Telco RAN DU reference design components

Ensure that components that are not included in the telco RAN DU profile do not affect the CPU resources allocated to workload applications.

Out of tree drivers are not supported.
Additional resources