There are different backup policies and characteristics on Triton's three data storage systems. This page describes the complete policies in effect at this time. If your project requires greater capacity or reliability than provided in this policy, please contact the Triton staff with details. Every effort will be made to accomodate your needs.

• Home Storage Area (Dual Copy Storage)

  Home areas on Triton are located on Solaris-based NFS servers using ZFS as the underlying file system. There are 36 terabytes of space in the Home Storage system for all users.

  • Local HD Redundancy: Double-drive failure can occur without data loss: Triton uses Raidz2 variant of Raid6
  • Connectivity: 10GbE; Delivers > 300MB/sec to single node; > 500MB/sec aggregate
  • Space: Each user is guaranteed at least 50GB of space
  • Backup/Replication: A snapshot of each user's home area is taken nightly and replicated to an identical server; snapshots are retained for at least 7 days (longer if space available); no other backup is performed; view the snapshot recovery page
  • Total Space for files, snapshots: 36TB
  • Location: Snapshots accessible at $HOME/.zfs/snapshot
• Lustre Storage Area (Single Copy Storage)

  Triton has a parallel file system (PFS) known as Data Oasis. It contains about 800 terabytes of shared scratch space available to all users.

  • Local HD Redundancy: Single-drive hardware failure is supported through Raid5 on the Lustre Object
  • Connectivity: 4 x 10GbE; delivers > 500MB/sec to single node; > 2.5GB/sec aggregate
  • Space: Each user is given access to this storage as part of their allocation; at this time, no minimum guarantee or maximum limit is defined; if a project needs medium term (1-6 mo.) data availability, Triton supports project special requests. Please contact Triton help.
  • Backup/Replication: No backup of this storage is performed
  • Total Space for files: 800TB
  • Location: Storage accessible at /phase1/<username>
  • Location: Job-specific storage accessible at /phase1/scratch/<username>
• Local Node Temporary Storage

  A small scratch space is available to all users. This comprises approximately 5 gigabytes per node of available storage only during job runs. It is purged between jobs.

  • Local HD Redundancy: Single-drive hardware failure is supported with Linux SW Raid-1 (Mirroring)
  • Connectivity: Local HD; about 50MB/sec/node; about 14GB/sec aggregate
  • Space: Generally about 5GB/node; purged between jobs
  • Backup/Replication: No backup of this storage is performed
  • Total Space for files, snapshots: Dependent on number of nodes requested for job
  • Location: Storage accessible at /tmp from local node only