This service is formerly known as Faculty Owned Equipment (FOE).
Advanced Research Computing (ARC) allows researchers to purchase research hardware to be integrated into our high-performance computing (HPC) clusters: Great Lakes, Armis2, and Lighthouse.
Using the pay-as-you-go options available on Great Lakes or Armis2 may be a more affordable choice, especially with cost-saving features like the no-cost allocations provided by the University of Michigan Research Computing Package (UMRCP), and any unit-specific support. Access to pay-as-you-go on Great Lakes and Armis2 can be set up quickly.
Purchasing and integrating custom equipment should be considered if:
- Required by a grant
- You have unique hardware requirements not offered by the existing pay-as-you-go services
- You need to ensure your research work will be completed in the window of the grant and the only way to ensure this is to have priority on the research hardware
Hardware Integration Models
Researchers can choose between two integration options when adding their hardware to our high-performance computing clusters.
Condo Model
The “condo model” gives researchers exclusive access to the hardware purchased via research funding. This means they can run jobs using research purchased resources, but if the hardware experiences a failure or goes down, there is no backup hardware available until the service contract for the hardware can restore it to operations, which can disrupt their research.
Benefits of the Condo Model
- Allows the PI to prioritize completing the work required by the grant on the grant funded hardware during the life of the grant
- Greater customization of cpu/memory/gpu and local working storage
Timeshare Model
The “timeshare model” allows researcher hardware to become part of an existing shared pool. This means that while researchers have priority access to their equipment, the hardware is available for use by others when not in use, significantly increasing overall access to computing resources for the community.
The flexibility of the timeshare model enables researchers to fully utilize the cluster’s resources, giving them access to a variety of computing environments based on their needs.
For the Great Lakes and Armis2 clusters, ARC will centralize on common hardware types to meet researcher needs and add that hardware to available partitions, ensuring efficient and standardized resource integration.
Benefits of the Timeshare Model
- Burst beyond the purchased capacity—up to 25% additional resources—whenever you need more computational power
- Flexibility to fully utilize the cluster’s resources
- Protection against hardware failures
- Receive credit proportional to the hardware investment that can be used anywhere on the cluster
- Hardware is used to its full capacity
- Increased computational resources for the U-M community
Researcher Integration Options by Cluster
| Cluster | Model Options | Existing Timeshare Partitions | Additional Notes |
|---|---|---|---|
| Great Lakes | Timeshare | standard, spgupu, spgpu2 | Researchers must leave hardware in the cluster for five years or the cluster’s lifetime, whichever is shorter. |
| Armis2 | Timeshare, Condo | Researcher slurm account | Condo model carries risk of no backup resources. |
| Lighthouse | Condo | Researcher slurm account | Condo model carries risk of no backup resources. |
Working With ARC
Researchers interested in the ARPH service must work with the ARC team to purchase and integrate their hardware. ARC will assist throughout the purchasing process and ensure smooth integration into the cluster.
Note: ARC cannot integrate hardware into the cluster that is purchased without our knowledge.
Budgeting for ARPH
ARPH billing includes a one-time hardware cost and ongoing hosting subscription costs that are determined based on the number of nodes purchased. The hardware and the subscription can be different sources.
All ARPH purchase and subscription costs are charged to a U-M Shortcode.
ARPH usage is billed monthly, with rates subject to change over time. The details of the charges will appear on the normal monthly Statement of Activity.
ARPH Agreement
For timeshare installation on the Great Lakes and Armis2 clusters, the ARPH agreement requires researchers to leave their hardware in the cluster for 5 years or the duration of the cluster's lifecycle, whichever is shorter, to ensure appropriate distribution of the provided credit.
Rates
For more information on the rates associated with the Researcher Owned Hardware service and other ARC services, please visit the ARC Service Rates page.