Uploaded image for project: 'SAFe Program'
  1. SAFe Program
  2. SP-1692

Establish test and benchmarking environment for SDP pipelines

Details

    • Spike
    • Must have
    • PI11
    • COM SDP SW
    • None
    • Data Processing
    • Hide

      In both SP-1548 and SP-1585 we have made a start in defining tests of SDP pipelines, yet so far we are lacking a standard environment to run them in. This is needed, as

      • developers shouldn't have to worry about acquiring an environment for testing their pipelines
      • we need to make sure we compare apples to apples for tests and benchmarks - we need a "gold standard" here
      • and finally - viewed from the other perspective - we need to gather experience in designing, establishing and running (small) HPC-like systems for our pipelines to inform the architecture going forward.
      Show
      In both SP-1548 and SP-1585 we have made a start in defining tests of SDP pipelines, yet so far we are lacking a standard environment to run them in. This is needed, as developers shouldn't have to worry about acquiring an environment for testing their pipelines we need to make sure we compare apples to apples for tests and benchmarks - we need a "gold standard" here and finally - viewed from the other perspective - we need to gather experience in designing, establishing and running (small) HPC-like systems for our pipelines to inform the architecture going forward.
    • Hide
      • Elaborated environment design (https://confluence.skatelescope.org/pages/viewpage.action?pageId=142968310)
      • Especially make decision concerning workload scheduling (concrete technology, scheduling considerations)
        • Show that CI jobs can be triggered through the workload scheduler from GitLab, with useful reporting. Ideally a distributed pipeline developed in SP-1548 or SP-1585.
        • Stretch: Look into how to interface with SKA LMC (i.e. implement SLURM or Nomad deployer component as alternative to Helm?)
      • Implement design (proof-of-concept depending on availability)
      • Socialise results to other teams
      Show
      Elaborated environment design ( https://confluence.skatelescope.org/pages/viewpage.action?pageId=142968310 ) Especially make decision concerning workload scheduling (concrete technology, scheduling considerations) Show that CI jobs can be triggered through the workload scheduler from GitLab, with useful reporting. Ideally a distributed pipeline developed in SP-1548 or SP-1585 . Stretch: Look into how to interface with SKA LMC (i.e. implement SLURM or Nomad deployer component as alternative to Helm?) Implement design (proof-of-concept depending on availability) Socialise results to other teams
    • 3
    • 3
    • 20
    • Team_PLANET
    • Sprint 5
    • Hide

      Not finished due to people leaving?

      Show
      Not finished due to people leaving?
    • 14.2
    • Outcomes Reviewed

    Description

      This is following on from SP-1548 and SP-1585:  We need to ensure that we actually have a stable environment to run them on. This will likely require quite a bit of interaction with different teams (system, pipelines, external) and is also subject to availability of hardware, so we should be careful about what exactly we commit to here.

      Attachments

        Issue Links

          Structure

            Activity

              People

                p.wortmann Wortmann, Peter
                f.graser Graser, Ferdl
                Votes:
                0 Vote for this issue
                Watchers:
                2 Start watching this issue

                Feature Progress

                  Story Point Burn-up: (82.98%)

                  Feature Estimate: 3.0

                  IssuesStory Points
                  To Do00.0
                  In Progress   14.0
                  Complete919.5
                  Total1023.5

                  Dates

                    Created:
                    Updated:
                    Resolved:

                    Structure Helper Panel