Hive

Hive is a centrally managed cluster with standardized hardware and connectivity and a defined life cycle of support. Hive was built to provide greater access to HPC for a larger number of users on campus while maintaining support for college-level needs. HPC@UCD offers two tiers of support moving forward and incentives to merge existing hardware when possible.

Hive Hardware

Hive consists of 14 nodes, each with 128 CPU cores and 2 TB of RAM. There are 4x NVIDIA A6000s for general use, and a single 4x NVIDIA H100 node. Nodes that are new enough are being migrated from other campus clusters to simplify administration and maintenance.

Hive uses a parallel file system Quobyte for storage. This is a very exciting change, and a first for any cluster at UC Davis. A parallel file system allows performance and capacity to grow over time, keeping up with research demands, all while keeping a single directory for PIs.

Access to Hive

UC Davis staff, faculty, and graduate students are entitled to free access to:

• Partition high: the entire group has combined access to 128 CPUs, 2000 GB of RAM, and 4 GPUs (NVIDIA A6000s).

  • Each job is limited to a maximum of:

    • 8 CPUs
    • 128 GB of RAM
    • 1 GPU

  • You can access these resources by adding --account=publicgrp --partition=high to your srun or sbatch command.

• Partition low: This partition allows access to all the unused resources in the cluster. The downside is that when a job is submitted to high that needs these resources, your job will be killed and requeued. Unless your job can do automatic check-pointing and restarting, all progress is lost when the job is killed.

  • Each job is limited to a maximum of:

    • 3 days (3-00) of runtime
    • 1 node

  • You can access these resources by adding --account=publicgrp --partition=low to your srun or sbatch command.

In addition to this, each new user is allocated a 20GB home directory.

To request free access, follow the request a new account documentation and request access to the UCD HPC Sponsored Public Access (publicgrp) from the list of sponsors through Hippo.

Additional resources on Hive may be purchase two ways. Sponsors (PIs and group leads) may purchase individual compute cores and Tb of storage at our current rates. In consultation with HPC@UCD, sponsors may also purchase approved nodes to contribute to Hive. HPC@UCD will work with sponsors to understand budgets and specifications and request quotes.

Contributors always receive priority access to the resources that they have purchased within one minute with the "one-minute guarantee" through the high partition.

Finally, Hive has a node available for use in graduate teaching. Email hpc-help@ucdavis.edu to request information or access to these resources.

Hive Administration

Hive hardware and software are administrated by the HPC@UCD team.

Current Rates

Rates for Hive can be found here.

For purchases and inquiries, please contact hpc-help@ucdavis.edu.