Step-by-step guide

The following step-by-step guide describes a typical scenario for HEAppE deployment at IT4Innovations National Supercomputing Center:

  • Alice, Bob, and Arnold are three researchers/developers of application/code. Their application processes a large amount of data and they wish to both reduce the computational time and decrease the output waiting time. This speeding-up of their application requires resources that are beyond the scope of conventional hardware/software. So, they decide that they would like to utilize the HPC cluster at IT4Innovations.

  • Computation resources. There is a necessity to have a project for computational resources. They decided to ask for computational resources via Open Access Competition for 500 node-hours (link). The grant competition is announced 3 times a year (February, June, and October).

  • They filled out the application form and submitted it via the web portal (link). Alice and Bob agreed the prime investigator (PI) is Arnold.

  • In order to submit the application form they need to gain access to IT4Innovations’ Information systems and obtain project membership (link). As the PI, Arnold creates an account for communication with IT4I support and for accessing the SCS information system (having an account does not mean that you have immediate access to IT4I clusters).

  • Congratulations! Their project has been evaluated and approved by external reviewers.

  • They then receive a project for computational resources with the ID OPEN-XX-XX. Their IT4I account is assigned to a project, and they can access the IT4I clusters.

  • Access to IT4I HPC clusters. There are two options: direct or indirect (through Middleware) connection. Since they do not have any experience in the HPC domain and do not want to run their jobs straight on an HPC cluster they decide to use the indirect HEAppE Middleware technique.

  • There are two ways to deploy the HEAppE Middleware: manual deployment (option A), and through a HEAppE supported deployment (option B); a combination of the two is also possible. Because manual deployment is proven to be slower and riskier, it is recommended to use option B, and therefore to create a high-confidence deployment with full control over it. Consequently, they pass on the responsibility of deployment to the HEAppE/SCS support.

  • The HEAppE support then sends functional account requirements (a default of five) to the SCS support. Arnold, as the project-responsible person (the PI), must approve all requirements (usually via email).

  • After Arnold’s approval, the HEAppE support can create an instance for his project (OPEN-XX-XX).

  • External HEAppE accounts are created for all three researchers (Alice, Bob, and Arnold) to be able to connect from outside of the network. Consequently, admins send credentials to all participants via a secured connection.

  • Optional: Arnold can ask for extra (no default) HEAppE specifications (e.g. command template, link).

  • It is also necessary to prepare and upload the research application (its source code) in the project storage (which Arnold does as he is the PI).

  • Customized template creation. Command templates initiate commands to create one or more items programmatically. To create the HEAppE command template, the support needs the primary users’ preferences such as usage of CPU/GPU or specific queue for queueing the jobs.

  • Finally, we are finished, they can explore their application and data via the HEAppE Middleware (and run an HPC job). They can use HPC power without a direct connection to any IT4I clusters.