On-Prem vs. Cloud

Carina is Stanford University’s on-prem computing environment for high-risk data. Carina On-Prem leverages the Stanford Research Computing private cloud. The Nero GCP environment is Stanford's cloud service for high risk data, built on Google Cloud Platform.

Please take a look at our overview of services below. Our team remains available via email at srcc-support@stanford.edu and via office hours.

	Carina On-Prem	Nero GCP
Pre-Requisites	Project led by a Stanford PI Users must have a full Stanford affiliation Completed Data Risk Assessment	Project led by a Stanford PI Users must have a full Stanford affiliation Completed Data Risk Assessment A valid PTA
Application and Services	Slurm JuypterHub NoVNC	JuypterHub Native Google Cloud services like BigQuery, Compute, Storage, Vertex AI, Kubernetes Engine, etc.
Analytical Tools	No difference between Carina and Nero GCP. Analytical software: Python, R, RStudio, SAS, Stata (with proof of existing license) Package management tools: Anaconda (Python and R), Pip (Python), Lmod Source code management tools: GitLab
Application Access Methods	SSH: Slurm Interactive: JupyterHub, NoVNC	SSH: Command Line Interface Interactive: JupyterHub, GCP console, or gcloud command to native GCP services Programmatic: Restful APIs and Client lib for native GCP Services
Storage	Storage Type: File system Base Allocation: Home Directory: 15 GB /home/$SUNETID PI Directory: 1 TB /share/pi/$PISUNETID Additional storage can be ordered at a rate of $91/month for 10TB.	Storage Type: Object Storage (GCS), File system (Filestore), Data warehouse (BigQuery), and RDBMS (Cloud SQL) Base Allocation: ($) Unlimited, cost quotas can be set up based on user preference.
Data Transfer Methods	SFTP (Secure File Transfer Protocol): Cyberduck (interactive), SFTP command (Terminal) Rsync/rclone gsutil rsync and gsutil cp: Command line utility, move data from on-prem to GCS, or GCS to on-prem.	gsutil rsync and gsutil cp: Command line utility, move data from on-prem to GCS, AWS to GCS or GCS to GCS. Handles medium and small size transfer, TSOP (Storage Transfer Service): Managed service, move data from on-prem to GCS, AWS to GCS, GCS to GCS, and file-system to file-system. Handles large volume transfer (>1TB) smoothly
Compute Options	Login node: 1 Compute nodes: 16 - Dual Core Intel Xeon Gold 6130 (16C 2.1GHz), 384 GB RAM GPU nodes: 14 - (2) Dual Core Intel Xeon Gold 6130 (16C 2.1GHz), 384 GB RAM + Dual P100 GPUs (6) Dual Intel Xeon Silver 4114 (10C 2.2GHz), 256 GB RAM, Quad NVIDIA Tesla V100 (6) Dual Core Intel Xeon Gold 6330 (28C 2GHz), 512 GB RAM, Quad NVIDIA A100 (40GB) OS - Ubuntu 20.04	($) Scalable depending on research needs and budget Custom machine size to meet research needs. A large variety of machine types to choose from: general purpose, compute-optimized, memory-optimized, accelerated optimized. GPU: A100, T4, V100, P100, P4, K80 TPU: V2-8/512, V3-8/32
Security and Compliance	Is compliant with Stanford’s minimum security requirement for high-risk data Is covered under Data Risk Assessment #1479	Is compliant with Stanford’s minimum security requirement for high-risk data Is covered under Data Risk Assessment #1479 Certified Platform for High Risk Data, HIPAA Compliant, Covered under the Google-Stanford BAA. Please check out GCP Compliance Resource Center for covered products and services.
What to Expect	Cost: Fully subsidized as of 2024 Experience: Access to a shared cluster with a variety of compute options Compute experience offers interactive and SSH options, with job scheduler (via Slurm) available.	Cost: Explicit GCP costs with negotiated discount, pass through to faculty via a Stanford Billing Account, called a PTA number. Experience: Scalable Unified console to access a rich set of hardwares and services
How to Get Started	High Level Steps: Complete User prerequisites (DRA) Request a Carina On-prem account Connect to Carina On-Prem	High Level Steps: Complete User prerequisites (DRA) Request a Nero GCP project Connect to your Nero GCP Project Access your application and native GCP Services
Support	Available Options: SRCC Team: account and environment provisioning, office hours for onboarding and troubleshooting Stanford SRCC Slack Channels: #carina-announce, #carina-users	Available Options: SRCC Team: project and environment provisioning, office hours for onboarding and troubleshooting Stanford SRCC Slack Channels: #nero-announce, #nero-users GCP Technical Support: technical issue with GCP products and services, suggestions, feature request GCP Account Team for Stanford: architecture mapping, use case discussion, cost estimation, enablement, support ticket escalation, early access to new features, office hours. GCP Slack Channel @Stanford: #googlecloudplatform