Child pages
  • ROGER - User Policies
Skip to end of metadata
Go to start of metadata

NOTE: This policy is currently in draft stage

Reporting

  • NCSA will be reporting to NSF the usage of ROGER by users.

    • Users are responsible for reporting their work back to NCSA.

    • Acknowledgment of ROGER system use and the NSF grant that supports it (NSF-1429699) must be made in, proposals and publications, trainings and workshops that used ROGER, any data products, student engagement (including classes), and science impact stories.

    • We may also conduct periodic usage surveys or other means of gathering information.

  • NCSA will report the following to NSF:

    • Number of jobs run

    • Jobs run per queue: High memory, Hadoop, etc.

    • Data movement

    • Point-in-time usage

    • Project allocation

    • Tickets opened/closed and duration

    • Portal data

    • OpenStack data

ROGER Filesystem

  • NCSA will provide filesystems based on GPFS and provide partitions suitable for  small and large files. Currently 3 partitions are provided /home, /project and /scratch. All users are given a home, project and scratch directories. The project and scratch quotas will be associated with the project while home quota is individual based. If more space is required a ticket needs to be initiated by the project.

  • A purge policy will be investigated in the near future. Scratch will be purged on a 30-day last-access date same as Illinois Campus Cluster Program (ICCP).

  • Hadoop HDFS is required to be local on the solid states per Hadoop node and it is the user’s responsibility to copy any data to be saved from node to scratch or other file systems as needed and before job ends.

Change controls

  • The project understands that there are times when changes need to occur on the system and maintenance will need to be scheduled. Machine changes, downtime and maintenance will be announced via email on ncsagroup+prj_cg_users@ncsa.illinois.edu

Help desk

Security

  • Need to conform to NCSA security policy.
    • [More Info needed ]

Resource configuration 

  • There are 3 different types of configurations: batch compute, Hadoop, and OpenStack.
    • The batch compute is available to all users.
    • Hadoop and OpenStack are by special request only via the ticketing system.
    • The policies of OpenStack are currently being refined.
  • All resources are contingent upon an Allocation process.
  • The compute resources can be rearranged as needed.

Queue system

  • There are queues available for users to submit jobs per batch.

    • The queues available are development and normal with flags available for high memory and GPU.

    • Reservations will be placed as needed, please submit a special request via the ticketing system.

  • Hadoop queue will be Yarn based.

    • Batch jobs will be run via one node only, they will not share resources.

Storage allocation

  • Home quota: 10GB/user

  • Scratch quota: 2TB/project

  • Project quota: 500GB/project

    • More can be requested, project directories should include HDFS and VMs. If more needs arrive, please submit a ticket. NCSA is providing this space for longer-term requirements.

    • Grace period of 30 days past allocation. Possibly up the project space (not the home).

    • Project and scratch space are organizationally up to the PI.

    • Notification of overages will be sent via email.

    • Hard and soft quotas exist; writing can continue until hard quota is reached. ACTION also add to allocation email and policy

Backups

    • Once-a-day per directory snapshots are kept for two weeks. If the file system is lost, the back ups are lost. NCSA also supplies and off-site environment if a second copy doesn’t exist for the project already. ACTION put in user guide that data is vulnerable to a disaster or file system loss. Projects are responsible for their own disaster recovery plan.

 

  • Data management system

    • NCSA will provide the Globus online infrastructure for data management. This provides GridFTP, gsiSSH. These will run on separate nodes with 40gb ethernet for increased performance. Other data movement utilities can be run from the login nodes as individual users; such as, RSH, SSH, FTP. Large data transfers are recommended through Globus Online.

 

  • Software

    • NCSA provides the following software--point to document. NCSA provides the operating system and the tools that are required by the project. System administrators will install tools as needed via ticketing system. Additions or security updates will occur over time and will be coordinated with project office in due time. If additional software is needed please submit a ticket. PI will review software/administration mechanism (PI vs. Project expense) on a case by case basis. ACTION JD make this a living document with references on ROGER wiki

 

  • Allocation process

    • NCSA will allocate users and storage based on peer review, needs, and requests. This is a shared resource, so some resources may be limited due to availability. The original PI partners will offer 6 month submission windows. Any shared partners will need to submit usage plans that will be evaluated based upon availability. The allocation process will be via committee and subject to annual review.

  • No labels