Build HPC cluster placement group configurations with RDMA networking and cluster network settings.
Last verified: May 2026
Build HPC cluster placement group configurations with RDMA networking and cluster network settings.
Required Fields
compartmentIddisplayNameavailabilityDomainclusterPlacementGroupTypeOutput will appear here...The builder constructs OCI Cluster Placement Group configurations: placement group resource (compartment, name, AD), instance launch placement constraints (group OCID + RDMA-capable shape + cluster network selection), and capacity reservation linking for guaranteed availability. Output is generated as oci compute cluster-network commands and Terraform oci_core_cluster_network + oci_core_compute_cluster resources.
Build HPC cluster placement group configurations with RDMA networking and cluster network settings. This tool helps OCI engineers generate valid configurations quickly without consulting documentation, reducing errors and accelerating infrastructure deployment. All processing runs in your browser with no data sent to external servers.
Your ML team is training a large language model that requires 64 H100 GPUs with low-latency interconnect. Without cluster placement, GPUs scattered across racks would have ~milliseconds of inter-node latency — devastating for distributed training. The builder generates a cluster placement group with BM.GPU.H100.8 shapes, RDMA cluster network, capacity reservation for guaranteed availability. Training time drops from projected 4 weeks (with high latency) to 5 days (with RDMA). Saves $40K in compute cost on a single training run.
Cluster placement groups + RDMA networking are essential for tightly-coupled HPC workloads (MPI, GPU training across nodes, high-performance simulations). Without placement groups, instances can be scheduled anywhere in the AD; with them, instances are guaranteed to be on the same physical rack.
RDMA-enabled shapes (BM.HPC2.36, BM.GPU.A10.x, etc.) provide microsecond-level latency between nodes via cluster network. For ML training with thousands of GPUs, this is the difference between a 1-day training run and a 1-week training run.
Cluster placement groups have capacity limits — you can't always get all the nodes you need on the same rack. For workloads needing 100+ tightly-coupled instances, capacity reservations are essential to guarantee availability.
Most Cluster Placement Group primitives behave the same in commercial and Government Cloud OCI, but the OCID realm differs, region availability is limited, and a handful of services are unavailable. The output is portable in shape; you must adjust realm and verify service availability before applying in a Government Cloud tenancy.
The Cluster Placement Group options surface what is currently documented in the OCI reference for that service. When Oracle adds a new property or value, we add it here after verifying the schema in a real tenancy. If a recently-announced feature is not yet selectable, treat that as a 'not yet supported' signal rather than an opinion that it should not be used.
Was this tool helpful?
Disclaimer: This tool runs entirely in your browser. No data is sent to our servers. Always verify outputs before using them in production. AWS, Azure, and GCP are trademarks of their respective owners.