Multi-instance vs Single Instance

There are two confirguation options for Nero GCP projects, review the chart below to identify which option is best for your project.


Overview	Kubernetes deployment with shared Jupyter access and storage for the whole team.	Each compute user gets a dedicated instance.
Ideal For	Team needs a shared, always-on environment Different people need different instance sizes on demand	Team members mainly work independently Team computing needs do not warrant the expense of a always-on enviroment
Cost	~$300/month fixed cost to run the cluster 24/7, plus usage-based compute charges	Billed only for actual compute usage and minimal persistent disk costs
Auto shutdown	Inactive sessions stop automatically	No automatic stop — users must remember to shut down manually
Access	One shared web address with a spawn menu to launch different instance sizes	Individual instance per user — access via gcloud SSH with web interactivity (with SSH and port forwarding)
Software	JupyterLab 3.2.8, RStudio 4.4.2, SAS 9.4	JupyterLab 1.2.16, RStudio 4.4.3, SAS 9.4, Stata/SE 16.1
Storage	Shared PI directory (/share/pi/$sunet-PI) for team files, plus GCS bucket access	100GB boot + 100GB /home/jupyter per instance; plus GCS bucket access

Note: Teams aren’t locked in — you can switch between a multi-instance cluster and single instances later if one doesn’t fit your workflow

If you have additional questions email us at srcc-support@stanford.edu(link sends email) or sign up for office hours.