Wynton Transition Project FAQs

What is the timeline for transitioning data from Wynton to Core HPC and FAC storage?

How do I tag directories in Starfish, and when can I start?

As of December 19, 2025, the Wynton Protected migration process has changed, and Starfish will no longer be used to tag directories.

What storage options are available on Core HPC?

- Capacity Storage: Use for large datasets; includes backups. 
- Home: Limited, fast storage for scripts. 
- Scratch Space: Temporary storage for computation, with a maximum storage duration of 30 days.  

Do I need to purchase FAC storage to use Core HPC?

Yes, Core HPC does not have persistent storage of its own; project data must reside on FAC storage. 

How can I estimate the cost of storage?

Use the Storage Cost Calculator, which accounts for storage costs, backups, and service fees.

Will credits from Wynton contributions be transferred to FAC storage?

Yes, credits for contributions will be transferred based on depreciation schedules (typically five years). The Wynton Transition team will reach out to eligible users.

What options do PIs have if the credit memo does not meet their needs?

We recognize that the Wynton migration may be disruptive for some research groups, particularly those with specialized or long-term compute needs. While the credit memo provides one pathway, it is not the only option available.

PIs who receive a credit memo have two additional options to consider:

1. Hardware Withdrawal (ITOM Model 3 – Independent Research Computing)

PIs may request to remove their existing hardware from Wynton and operate it independently under ITOM Model 3.

Model 3 is designed to support research flexibility and innovation, allowing labs to run their own systems while maintaining institutional standards. However, this option comes with important responsibilities and guardrails.

Key requirements for Model 3:

• Systems must comply with UCSF Minimum Security Standards (https://tiny.ucsf.edu/mss)  
• Systems must be inventoried in ServiceNow (CMDB) and have clearly defined ownership
• Required security controls must be in place, including:
  • Endpoint security (e.g., EDR/antivirus/firewall)
  • Vulnerability remediation within defined timelines
  • Encryption (at rest and in transit)
  • Strong access controls (including MFA for administrative access; no shared credentials)
• Systems must operate within laboratory infrastructure limits (power, cooling, safety) 

Data classification guardrails:

• P1 (Public) and P2 (Internal) data: Allowed under Model 3 if all security standards are met
• P3 (Sensitive) data: Requires formal approval via Model 3 attestation
• P4 (Restricted) data: Not permitted under Model 3 (must be enterprise-hosted) 

Additional considerations:

• Labs are responsible for ensuring compliance with grant requirements, export controls, and institutional policies
• Systems that cannot meet minimum security standards must be remediated or receive a formal security exception
• Laboratory environments must be able to support the hardware (power, cooling, and safety constraints apply) 

Model 3 enables continued use of existing hardware, but does not reduce institutional security or compliance obligations.

2. Financial Assistance

PIs may also reach out to us to explore potential financial support from the university to help offset the transition to alternative compute resources (e.g., coreHPC or other solutions).

Support availability and structure may vary, but we strongly encourage faculty to initiate this conversation if the credit model is not viable. 

Why might these alternatives be important?

For some workloads, especially those requiring multi-GPU configurations or large attached storage (e.g., cryoEM), the credit model may not adequately reflect the cost or capability of existing systems, and having access to the latest hardware (as is set up in FAC and CoreHPC) may not be important.

For example:

• A PI with older GPU nodes (such as A40 GPUs, which have high VRAM but are a couple of generations behind) may receive a one-time credit that does not enable an equivalent compute capacity  
• Comparable systems through CoreHPC may involve significantly higher upfront and ongoing costs, while offering reduced GPU density
• Additional requirements such as a need for large storage capacity (hundreds of TBs) further increase total annual costs 

In some cases, the credited amount would force one to move from multi-GPU nodes to significantly reduced compute capacity with added annual costs, making the credit option impractical.

What should I do if I want to explore these options?

If you are considering Model 3 or would like to discuss financial assistance:

• Contact the Wynton transition team at [email protected]
• Include a brief description of your current hardware, workloads, and storage needs 

We strongly encourage PIs to reach out early so we can help identify the most appropriate path forward.

Where can I find more information and updates?

A dedicated Wynton Transition page has links to many resources . Visit the CoreHPC webpage for details about features, specifications, and updates.

Where can I get help with the transition or ask questions?

You can email [email protected] for general help. Please refer to the Wynton Transition page for additional support options.