The Unity HPC and AI Platform is available to academic institutions through our Unity Open Access program. Academic institutions lease CPU, GPU, and storage resources on an annual basis and make those available to their users.
Rates
The table below reflects standard annual Unity node configurations and rates. Specialized hardware (for example, high memory nodes, non-x86_64 architectures, or particular GPUs) is available. Please inquire for more information at hpc@umass.edu.
| Resource | Annual Rate |
|---|---|
| 128 Core Compute Node (1T RAM, InfiniBand) | $11,848.83 |
| GPUs (A40)* | $3,951.70 |
/work storage per TB | $345.42 |
/project storage per TB | $56.55 |
* Other GPU pricing available upon request, email hpc@umass.edu for information.
Unity provides two expandable storage platforms:
/project directories (intended for storage and not to be used for running
jobs) and VAST-based /work directories (intended for active data).
Access
Our systems administrators work with institutions to determine appropriate access policies for their affiliated researchers.
Support
Rates include Tier 1 and Tier 2 support. Tier 1 support includes access to email-based ticketing and the Unity User Community Slack support. Tier 2 support includes modest individualized consultations with researchers as needed. Tier 3 support is available on an hourly basis for research groups needing in-depth consultation and development services.
Standard resource allocations
Research groups at partner institutions receive a standard storage allocation and compute access, as part of their institution’s Unity membership.
Standard storage allocations
Each research group receives access to storage, including:
- A 50 GB home directory for each group member.
- 1 TB of high performance work storage, located at
/work/<group name>. - Up to 5 TB of project storage, located at
/project/<group name>. - Access to temporary, high performance scratch space.
Standard compute access
Research groups can use the general access partitions,
including cpu, cpu-preempt, gpu, and gpu-preempt. Each research
group is restricted to using 1000 simultaneous CPU cores or 64 simultaneous
GPUs. However, there is no total core-hour or gpu-hour usage cap. At this time,
this restriction does not apply to the preempt partitions. For information about
submitting workloads to compute nodes, see our Slurm job scheduler
documentation. Note that availability of
resources is subject to cluster load and immediate access is not guaranteed.
In addition, all Unity users can access the Unity OnDemand portal, a web-based, graphical interface where users can access graphical applications; like JupyterLab, RStudio, and Matlab; and a graphical XFCE desktop.
Research group management
Upon account creation, we generate a unique
group for each eligible researcher, usually in the form
pi_<lead researcher's Unity username>. Group owners can manage access for an
unlimited number of group members, including students and research staff,
through the Unity portal.
Grant and startup funded hardware
Individual research groups can add grant- and startup-funded hardware to Unity at cost, with no additional maintenance fees. The total amount of such hardware at a partnering institution is limited to 25% of the fully funded institutional commitment.
Hardware owners get priority access to their hardware. However, all purchased
hardware is added to the cpu-preempt or gpu-preempt partitions for use by
all Unity users. Priority jobs supersede jobs in the preempt partitions
following a two-hour grace period.
Hardware additions must be approved by the Unity leadership team in advance and are subject to hardware type restrictions. To discuss a hardware purchase for your research group, contact hpc@umass.edu.