HWNI Neuro Cluster Technical Specifications

Last modified:December 16, 2019

Location

The Neuro cluster is located in the UC Berkeley data center, a hardened facility that provides a precision environment for mission-critical servers and related equipment.

Compute nodes

  • provide over 500 (3Ghz) Intel CPU cores and over 4TB of DDR3 RAM
  • provide 8 Graphics processors, including Geforce Tesla V100, Tesla M2050, and Geforce GTX470 (for graphics rendering)

Data Storage

  • SATA Enterprise drives provide over a petabyte (PB) of usable disk space.
  • XFS 64-bit Journaling Filesystem is a stable, high-performance filesystem proficient at handling large files.
  • Network File System (NFS) Servers provide access to filesystems with multi-gigabit ethernet.
  • For increased availability, data backups are stored on a different file server (and filesystem) from the original data. For more information about data backups, see data backups.

Network Services

Network services include:

  • CUPS print service
  • Grid Engine batch processing service
  • Nagios monitoring service
  • LDAPS directory service
  • HTTPS web services
  • Secure login via 2-factor authentication ssh/sftp/sshfs

Network security

The network services listed above are deployed with virtual machines for increased redundancy, privacy and security. All login services use 2-factor authentication with SSH/SSL, restricted to pre-approved users with a strong password or SSH key. Repeated unsuccessful login attempts are logged and blocked. Workstations, NFS servers, and some virtual machines are only accessible from a private network address (RFC 1918) for increased privacy and security.

Software

The operating system is 64-bit CentOS linux.
Scientific Applications are located on workstations, updated nightly (via rsync) from a secure central repository.