Hardware Profiles

The CRCD operates several HPC clusters, each optimized for a different use case, plus the nodes that you connect through and the storage tiers that hold your data. You submit work to whichever cluster best fits your job. Choosing the appropriate one ensures that resources are used efficiently and reduces the time spent waiting in the queue. This section profiles each of them.

Compute Clusters

Submit jobs to these with the --cluster directive (or the matching crc-interactive flag). Each page lists the node hardware, partitions, and any group-specific investment hardware.

Cluster Best for
SMP Single-node jobs using shared-memory parallelism.
HTC Many independent single-node jobs (high-throughput); prioritized for NIH-funded and health sciences work.
MPI Tightly coupled multi-node jobs using the Message Passing Interface or similar frameworks.
GPU GPU-accelerated work such as AI, machine learning, and physics-based simulation and modeling.
Teach Instructional activities, such as coursework, tutorials and workshops, and hackathons.

Access & Auxiliary Nodes

These are entry points and special-purpose servers — not job-submission targets.

Node Purpose
Login The shared entry point for submitting jobs and managing files. Not for heavy computation.
Viz Interactive visualization and GUI applications on a Linux desktop.

Some partitions are investment hardware

A number of partitions are owned by specific research groups who invested in the hardware, and are restricted to those groups. Each cluster page notes its general-access partitions; see the Hardware Investing Policy for how investment works.

Storage

Compute is only half the picture — your data needs somewhere to live. CRCD offers tiered storage allocated per research group, from fast all-flash space for active jobs to low-cost archive. See Storage Tiers for what each tier costs and how to request more.

Choosing a cluster

Once you know which cluster fits, see Service Units for what a job costs, and the Job Scheduling Policy for per-group limits and how priority is computed. New to the clusters? Start with Getting Started.