Customer Profile
Indosat Ooredoo Hutchinson (IOH) is a leading telecom operator in Indonesia. Their market cap is roughly $4B and has 63 million cellular subscribers. IOH wants to monetize their edge sites for distributed inference and make them RAN-ready so that RAN workloads can be run alongside AI (AI-RAN concept).
Business Challenge
Needed to share GPU resources across multiple inference applications and RAN. Furthermore, IOH wishes to monetize any unused capacity through NVCF.
Solution
- Deployed Aarna Networks’ GPU Cluster Management Solution (CMS) to manage and orchestrate GPU resources at each edge site.
- GPU CMS automates the full lifecycle: server provisioning (Day-0), infrastructure setup (Day-1), and AI workload deployment (Day-2).
- Implemented EVPN-based multi-tenant network segmentation, enabling isolated GPU access for different teams or applications.
- Used NVIDIA GPU Operator to configure and manage MIG partitions and expose them as Kubernetes resources.
- Leveraged NVIDIA NVCF to enable cloud-native AI function deployment and inference.
- Enabled observability with built-in monitoring and logging for GPUs, apps, and infrastructure health.
- Provided a self-service portal with RBAC and policy control for secure, scalable access across teams.
- Orchestration of RAN DU/CU along with configuration of the front-haul switch, PTP, and RAN SMO
- Continuous dynamic scaling of RAN, AI, and NVCF tenants depending on the loading
Outcome
Demonstrated AI workload deployment via NVCF; future-ready architecture for RAN integration and multi-tenant scaling.
