Site maintain:Yu-Heng Tseng
Site credits

BAYWULF Policies

  • Machine access and user accounts
    • sshThe machine has 20 total nodes. Access is granted via ssh only to master1. Please develop and test your codes on master1 (baywulf.stanford.edu). Jobs can be run on master1 without the queueing system.
    • Home directories are located on master2. All home directories are mounted via nfs from master2.
    • Passwords can be changed with yppasswd to update the NIS database which is on master2. Using passwd will not result in a password change.
    • Accounts can be obtained by sending an email to admin stating the following:
      • Name
      • Preferred userid
      • email address
      • Machine name and IP address from which you will be accessing the cluster via ssh.
      • Brief description of the research you will be performing and the associated computational needs, including approximate memory/processor/disk usage and cpu time.
      • Account number to which we can charge maintenance. Your share of the maintenance costs will be computed by multiplying your fraction of total cpu usage by the total monthly mainenance costs. You will be informed of the charges when they are incurred.

  • Disk quotas
    • Users are given an initial quota of 5 Gb. Please send requests for more space to: space

  • Job size and priorities (currently not implemented) Submit jobs with PBS to:

    • pque: Parallel Queue
      • Priority: All users.
      • Nodes: 5-20
      • Maximum job size: 1 Gb per node.
      • Maximum job time: 24 hr.
    • psque:
      • Priority: Second to pque
      • Nodes: 5-20
      • Maximum job size: 1 Gb per node.
      • Maximum job time: 24 hr.
    • sque:
      • Priority: By application only. Please send your request to: request
      • Nodes: 1-4
      • Maximum job size: 2 Gb per node.
      • Maximum job time: 24 hr.

  • Backups
    • Scratch disk: All programs (serial and parallel) must write to the local /usr/scratch disks during execution. These can be found on all of the nodes. Please do not write to your home directory during execution because this will slow down the network connections for other users on other nodes. All data (not directories) in the /usr/scratch disks will be deleted on a weekly basis (except for master2:/usr/scratch, which is just a link to /usr/home/users).
    • Home directories: Full backups of home directories are performed on a weekly basis. Incremental backups are performed on a daily basis. Please email the system administrator for recovery inquiries: recovery