These are notes from a whiteboard discussion about the NDS Labs "big picture"
NDS Labs "Test Drive" | NDS Labs "Developer" | NDS Share "Staging" | NDS Share | |
---|---|---|---|---|
What is it used for? |
| <- Merge? Development Prototyping
| <- Merge? Alpha/beta testing |
|
When is it used? |
| Post-proposal development and integration | Alpha/beta testing | Production deployments |
What does it include? |
| Customized services Realistic but limited resources Fully-configured common services (e.g., iRODS federation, Shibboleth IdP, Swift storage) | Stable releases of integrated services Real users Real resources Real users Stable release | Actual scale resources SLA Security/Hardening |
Analogy |
| Development VMs | Staging system | Production system |
- Workbench/Test Drive: This is the current workbench service – supporting the ability to "test drive", demo, or trials services. This would be limited to services and configurations that are easily deployable via Kubernetes
- Developer: When users graduate from the Workbench, they are given access to real resources for development and integration. This might include services not easily available via Workbench such as a functioning iRODS preservation federation, Swift Object store, Shibboleth IdP, etc. Services wouldn't be started and stopped quickly – configuration options would be specified and services configured based on project requirements.
- Staging: When projects are ready for alpha/beta/rc testing, they would deploy their services to a staging environment. This would be limited-access, but on real resources (i.e., actual production configuration, but maybe not scale)
- Share: Once a project is ready for production deployment, it is moved to the "Share" environment. This could be hosted at AWS, GCE, TACC/SDSC/PSC/NCSA.
A note about Nebula
We spoke with the Nebula team about deploying "production" systems (services that need SLA/uptime guarantees). Nebula is currently not ready – for example, one compute node is failing, there's no UPS, the backend isn't parallel. Production services are currently running on VSphere/VMWare