Prepare the PSS software for deployment and testing at the Digital Signal PSI

https://miro.com/app/board/uXjVN7ik-wg=/?moveToWidget=3458764575319098336&cot=14

Who?

• PSS developers; CSP integrators; AIV engineers

Why?

• PSS have a remotely managed development environment with all the necessary hardware that will allow testing of the pipeline with different configurations (utilising the CPU/GPU/FGPA), in order to inform (a) the AA2 hardware, (b) the parametric model that will be used to optimise the performance of the pipeline, and (c) the deployment strategy.

What?

• The procured PSS server for development and testing is connected to the network at the TOPIC facility, and a VPN server is set up, and access to Team PSS is provided. The PSS software is installed on the PSS server and a suite of low-level hardware-validation tests are run, as well as a number of software tests, including that the PSS software can be directly compiled and run on the new server, that test vectors can be pulled from the PSS repository and ProTest can run on them, and the new server can be set up as a Gitlab runner.
• The PSS server is set up to perform benchmarking of the PSS pipeline, in order to inform and compare to the PSS Parametric Occupancy Model for SPS
• The PSS software deployment tools are trialed with the new server, and TOPIC collaborate with team PSS on improving those tools

Acceptance Criteria

• The PSS CI pipeline can run on the PSS server at TOPIC server (or on an existing PSS server while the PSS server at TOPIC is being procured) and can push PSS artefacts to CAR
• The PSS software can be deployed by an external user to the PSS server at TOPIC. The documentation should be sufficiently clear and easily discoverable to enable a user to do this.
• An external user can deploy PSS software builds for CPU/GPU from CAR and successfully run PSS Testing Framework on the TOPIC PSS server (sufficient user documentation exists to be able to do this)
• Limited demonstration of the PSS Parametric Occupancy Model. A demonstration of its principles of operation (e.g. how parallel/serial processes are represented and monitored in the model) is presented, with a focus on the Single Pulse Search.

––– assuming that the PSS servers have been set up at TOPIC –––

• Stretch: summarise outcomes of deployment at TOPIC and suggestions for improvement
• Stretch: Gitlab CI can be executed on a TOPIC PSS server