Skip to content

projectEureka is a web portal that runs a team’s apps, flows, and Slurm batch jobs on cloud machines, organized by project. You use it in a browser. The machines are started when you need them and removed when the work ends.

Platform Machines
AWS EC2 virtual machines, started from a launch template in an Auto Scaling group
Google Cloud Compute Engine virtual machines
Kubernetes KubeVirt virtual machines on your own cluster
  • Control node: runs the scheduler and stores the environment’s records (projects, users, launchers, flows, runs). It makes the cloud calls that start and stop machines, and it runs the language-model, quantum, and ML stages of a flow as cloud calls.
  • Login node: serves the web portal. It also hosts the Terminal (a shell in your browser) where you submit Slurm jobs, and it can serve the shared home and project directories.
  • Directory (LDAP): every user has a sign-in account and a home folder. A project can also have a project user, a Unix account with no sign-in. It is off by default and is created when an Owner turns on Project login. Its name starts with the Project user prefix (default _pe_). See Project settings.
  • Compute machines: started on demand for app sessions, flow stages, and Slurm jobs. Each one carries cost tags for the user, project, and environment. See How costs are tracked.
  • Storage: your home folder, the project directory (prj-storage), S3 or Cloud Storage buckets, and optional external NFS or OrangeFS mounts.
  • Outside services: language models (through Vertex AI, Bedrock, or an API key) and quantum services (IonQ, Amazon Braket).
  • On Kubernetes, quantum stages use IonQ with an IonQ API key set by an admin. Amazon Braket needs AWS, and ML training stages do not run on Kubernetes.
  • Retail prices are shown for AWS and Google Cloud machine types only.

Get started · Launch from a cloud marketplace · Roles and permissions · How flows run · Run Slurm jobs