HERD computing meeting



During the discussion several topics have been brought up:

  • Usage of one or more instances of RUCIO/FTS
    A single instance of RUCIO is the starting point for the prototyping work. A joint  task force will be set up to define the data management details. The usage of several FTS instances in order to avoid geographically distribute streamed copies as a fallback for third-party copies (tpc )seems not be necessary: should a tpc fail between two sites, a fallback strategy involving a third, geographically nearby data center can be set up to avoid involving the single FTS instance. The resulting single-instance FTS setup would be more easily manageable, and it is agreed that this solution will be the initial choice, to eventually be re-evaluated in future should any problem arise with it.
  • Usage of DIRAC as the job submission manager
    IHEP propose to use DIRAC, while INFN is currently developing a job submission infrastructure based exclusively on HTCondor. Technical details about the INFN solution will be provided to IHEP in a dedicated meeting.
  • Federation of the IAM deployments at CNAF and at IHEP
    IHEP needs an internal IAM instance for resources on private network. Federation with INFN IAM deployment should be investigated. For the time being the INFN IAM instance will be used for the RUCIO testbed to be deployed in China, similarly to what has been done for JUNO.
  • Evaluation of the cloud-based computing approach
    INFN will provide details about the containerized HTCondor working nodes running on Kubernetes and on the containerized job submission environment.
  • Next steps
    No following meeting has been scheduled; the IHEP and INFN teams will keep in contact via email for offline discussion and for calling meetings. A direct message system based on Matrix and developed by INFN will be evaluated for quick communications.
There are minutes attached to this event. Show them.