Crucible E-RAN: Quick Start Guide (Red Hat OpenShift)

Crucible E-RAN reduces RAN energy consumption by 10–15% using an AI world model of your network trained with reinforcement learning. It deploys as containers on Red Hat OpenShift, requires no hardware changes, and operates within your QoS guardrails.

Prerequisites

• Red Hat OpenShift 4.14 or later (validated versions are listed on our Red Hat Ecosystem Catalog page) with cluster-admin access to create a project
• Deployment bundle (Helm chart + container images) provided by Humanis during onboarding
• Helm 3.x and oc CLI on the installing workstation
• Persistent storage class (default sizing: 200 GiB) for telemetry and model state
• Network connectivity from the cluster to your OSS/counter source (Kafka topic, SFTP drop, or vendor OSS API endpoint — see the Deployment & Architecture guide)
• Optional, for on-cluster model training: NVIDIA GPU worker nodes with the GPU Operator installed (inference/recommendation mode runs on CPU)

The E-RAN images run as an arbitrary non-root UID and require no privileged security context; they run under OpenShift's default restricted-v2 SCC.

1. Install

Crucible E-RAN is deployed jointly with a Humanis engineer as part of your assessment or pilot engagement. Humanis supplies the container images and a Helm chart with a values file prepared for your network (vendor profile, ingest mode, site scope); installation into your cluster typically takes under an hour:

oc new-project humanis-eran
helm install eran ./eran-<version>.tgz \
  --namespace humanis-eran \
  -f values-<your-network>.yaml

2. Connect your network data

Edit the ingest configuration for your counter source and RAN vendor(s):

oc edit configmap eran-ingest -n humanis-eran

Set the vendor profile (nokia, ericsson, huawei, oran), the counter source endpoint, and the site list (or CSV upload via the dashboard). E-RAN begins building a calibrated world model of your network once counters flow; the calibration timeline is established during your assessment.

3. Set QoS guardrails

In the dashboard (oc get route eran-dashboard), open Settings → Guardrails and set the availability, latency, and coverage constraints the optimizer must respect. E-RAN never applies an action that its world model predicts would violate a guardrail.

4. Run in recommendation mode

E-RAN starts in recommendation (shadow) mode: it computes energy-saving actions and shows projected savings without touching the network. Review recommendations and projected vs. guardrail impact in the dashboard. This is the standard mode for assessments and pilots.

5. (Pilot/production) Enable closed-loop optimization

Once recommendations are validated on pilot sites, enable closed-loop execution per site group under Settings → Execution, connected to your OSS's configuration interface. Rollback is automatic if measured KPIs deviate from world-model predictions.

Verify the installation

oc get pods -n humanis-eran        # all pods Running/Ready
oc get route eran-dashboard        # dashboard reachable, shows ingest status "receiving"

Getting help

Support portal and contacts: /docs/support · contact@humanis.ai