YNHHS Hosted Resources

The Secure Aligned Flexible Environment (SAFE), deployed at Yale New Haven Health System (YNHHS), is a high-performance computing cluster specifically designed to enhance healthcare practice and research through extensive clinical data analysis. This platform provides a secure, compliant, and collaborative environment tailored for interdisciplinary teams to analyze a diverse array of clinical data sourced from the YNHHS EHR system. SAFE offers adaptable computing configurations that can be tailored to meet the complex computational demands of large-scale clinical and translational research, including the development of innovative medical AI applications.

The SAFE platform is ideally suited for experienced users who analyze large volumes of data, including images, text, and structured clinical data. It features a container architecture supported by the locally-developed Kamino interface, which allows users to perform specialized analyses using advanced algorithms implemented in Python or other packages. Additionally, the recent integration of GPUs into SAFE enhances its capabilities, facilitating cutting-edge medical AI research and development.

• 50+ nodes with 1,728 CPUs
• 32 GPUs (16 V100, 8 A100, and 8 H100)
• Over 1PB storage with integrated environments

Researchers at the Yale School of Medicine can request access to data and SAFE computing resources through the Helix Portal based on an approved IRB or Preparatory to Research (PTR) request. All members of the project team must be included on the IRB staff list of PTR ticket number. Upon approval, teams are granted a default computational quota, ensuring equitable access to computing power. Additional computing resources such as GPUs can be requested through the Kamino workspace.

Access to the SAFE Environment is provisioned with YNHHS Research Basic Access (RBA). The RBA request must be submitted by a faculty member in the covered components of the YU-YNHHS Covered Entity.

After you are connected to the YNHHS network, navigate to the Kamino login page. Important: This page is only accessible from the YNHH Network- you must either (a) be logged in to the VDI, or (b) be logged into the YNHHS VPN. See available environments and onboarding resources below.

SAFE uses the Kamino interface to provide containerized computing environments.

Available Environments:

• Jupyterlab for CPU with PySpark 3.5 – for use with Python
• RStudio (v4.5.1) with SparkR (v3.5.5) – for use with R

Note: Jupyter is no longer used for R. R users should use the RStudio environment.

Mapping Directories in RStudio

There is an additional step for users of R converting to RStudio. Your data will be found in /mnt/volumes/<your directory> which you will need to mount in order to easily access them.

Everything you have access to is available in /mnt/volumes. You should access this area for your data, team, and project information.

To mount your directories:

1. Click on the ellipsis (…) on the right side of the "Files" tab (see red arrow below):

2. Open /mnt/volumes
3. You should see your project folder and any data folders that were mounted

Given the special nature of the SAFE Environment in supporting electronic Protected Health Information (ePHI), NIH Controlled-Access Data, Controlled Unclassified Information (CUI), and other sensitive data types, usage accrues charges for all services. These rates reflect the additional cost of managing, supporting, and maintaining environments that meet stringent compliance requirements.

Compute Rates

**GPU charges are per hour of requested reservation, not per actual usage. Charges are multiplied by the number of GPUs reserved and the number of hours per day of reservation (1 day = 24 hours).

Storage Rates

Additional work-style storage beyond the included 1 TiB no-cost allocation is available at $5.15 per TiB per month. Storage charges are based on requested allocation, not actual usage.

The Kamino environment hosts a range of Llama models tailored to various NLP tasks such as text classifications, named entity recognition, text summarization and generation, and question-answering.

The YNHHS Research Virtual Desktop Infrastructure (VDI) allows remote access to a YNHHS secure desktop from any device, anywhere without using the YNHHS VPN. The VDIs are ideal for working with small-to-medium sized data sets, accessing Epic charts and Slicer-Dicer, and accessing the YNHHS-YU Computational Health Platform web client (Jupyter Notebooks and command prompts). Current software in the VDI includes SQL Server Management Studio, R, Stata, SAS, Office365, and Epic.

Access to the VDI comes with YNHHS Research Basic Access (RBA). The request must be submitted by a faculty member in the covered components of the YU-YNHHS Covered Entity. Upon submission, you will receive a welcome package with instructions on how to log into the VDI.