start-img Getting started

Using UNICORE

If you are an end-user or application developer who wishes to use an existing UNICORE installation, have a look at the user documentation for the UNICORE Commandline Client, the PyUNICORE client library or the REST API documentation.

Evaluating UNICORE

If you wish to experiment with a UNICORE server installation, one simple way is to download the Core Server Bundle which can be installed very quickly on a single test machine or even your laptop. See also Installing UNICORE for Testing for more information.

If you are a Docker user, we recommend using our UNICORE Docker image. for evaluation.

Deploying UNICORE

Full production deployments of UNICORE range from minimalistic to rather complex, depending on your requirements, use cases and existing infrastructure.

For the typical case of providing UNICORE access to a single compute cluster, please have a look at How to setup UNICORE for a single HPC cluster.

For more complex cases, here are a few starting points:

  1. For each target resource (e.g., a compute cluster), you need a TSI and a UNICORE/X.

    The TSI is deployed on the cluster’s login node(s). UNICORE/X, in contrast, should run on a separate server or virtual machine.

    UNICORE/X should not be run on a system where users can log in.

  2. We strongly recommend running a Gateway.

    One gateway is sufficient for all UNICORE services at an institution or company.

    This component shields backend services from direct external access, providing an important layer of security.

  3. For the services (except TSI, where this is optional), you will need server certificate(s) issued by a trusted Certificate Authority (CA), similar to what is used for web servers.

  4. For multi-site workflows, you will need a Registry and a Workflow service.