High Performance Computing
High End Visualization
About CCR Contact Us Events Facilities Faculty Project Highlights History Job Opportunities News Partners Staff
Bioinformatics Consulting Services Grid Computing Visualization
Accounts Application Software Consulting Getting Started Hardware Resources Frequently Asked Questions Machine Status MyStats Overview Request Help Storage Resources Training/Courses Web Portals
Access Grid Training Outreach
Events Logos Media News Photo Album Videos
Contact Information Directions
  Dashboard > CCR Web > User Overview > Lennon
Log In   View a printable version of the current page.
Lennon

CCR SGI Altix Hardware Page

Front-End (interactive use, debugging, cpu time limit of 30m)

  • Hostname = lennon.ccr.buffalo.edu (front end for logins, job submission)
  • Vendor = SGI
  • Architecture = Altix 350 Server
  • Number of Processors = 4
  • Operating System: Redhat Enterprise Linux 3.0 + SGI Propack 3.0
  • Processor Description:
    • 4 1.3GHz Itanium2 (Madison) Processors
    • Main memory size: 8 GBytes
    • Instruction cache size: 16 KBytes
    • Data cache size: 16 KBytes
    • Secondary unified instruction/data cache size: 256 KBytes
    • Level 3 Cache: 3MBytes

Production Machine (accessible only through the batch system)

  • Vendor = SGI
  • Architecture = Altix 3700
  • Number of Processors = 64
  • Operating System: Redhat Enterprise Linux 3.0 + SGI Propack 3.0
  • Processor Description:
    • 64 1.3GHz Itanium2 (Madison) Processors
    • Main memory size: 256 GBytes
    • Instruction cache size: 16 KBytes
    • Data cache size: 16 KBytes
    • Secondary unified instruction/data cache size: 256 KBytes
    • Level 3 Cache: 3MBytes

Processor Cache Performance Table

Property L1I L1D L2 L3
Size 16KB 16KB 256KB 3MB
Line Size 64KB 64KB 128KB 128KB
Ways Associative 4 4 8 24
Latency (cycles) 1 1 5-6 14
Bandwidth 48GB/s 24GB/s 48GB/s 48GB/s

Disk Layout

  • /home, User $HOME directories, mounted from the CCR SAN. Current disk quotas are 1GB.
  • /scratch, Primary high-performance scratch space, located on the Fibre Channel Disk array (10-20 times faster than USCSI). About 2.4 TB total.
    • Accessible through PBS, which will automatically create a directory in /scratch for each new batch job (the directory name will be /scratch/$PBS_JOBID).
    • Inside your PBS script, you can easily access the appropriate scratch directory using the $PBSTMPDIR environmental variable.
      • Always use $PBSTMPDIR to access your scratch space.
      • See the Sample PBS scripts for examples.
    • These directories should be cleaned up by each user, but they will be scrubbed automatically 4 days after the batch job ends.
    • Do not use /scratch to store your files! As a scratch space it will be scrubbed automatically. Files that need to be stored long term should be kept elsewhere.

Center for Computational Research - University at Buffalo - State University of New York