moe.met.fsu.edu home

Why "moe"?

Research Interests

Publications

Powerpoint Talks

Powerpoint Posters

Funding

Classes
  Met5534

Students
  Graduate
  Undergraduate
  Alumni

Web Pages

Software

Computer Cluster

Vita

Personal

Computer Cluster

---

Through support among several faculty members (Hart, Chassignet, Reasor, Erlebacher, Zou, Ruscher), our research group has access to a Linux Beowulf cluster designed by Hart, with gigabit internal networking.

You can view the status of the machines in the cluster at http://moe.met.fsu.edu/machines. You can view the status of the UPSs in the cluster at http://moe.met.fsu.edu/cgi-bin/multimon.cgi.

Hardware:

• 27 nodes
• 57 cpus
  • 56 >= 2.4ghz
  • 49 >= 2.8ghz
  • 25 >= 3.0ghz
  • 18 >= 3.2ghz
• Total memory: 41GB
• Total RAID5 disk space: 7TB
• Total unraided disk space: 5-10TB
• Tape drives: 100GB LTO1 on moe; 400GB LTO3 on otto

Software installed:

• OS: CentOS 4.6
• Portland Group Compilers (5.2-6.1)
• Intel Fortran Compilers
• GrADS (1.7-2.0alpha2)
• Matlab
• McIDAS-V
• Garp/GEMPAK (v 5.11.1)
• IDL
• Vis5D
• R
• mpich1
• wgrib, wgrib2, gribw, cnvgrib, grib2ctl.pl, g2ctl, NCAR spherical harmonic converters
• GoogleEarth
• NCAR Graphics & NCL
• WDSS2
• Solo
• Slicer
• netCDF
• FFTW
• Many other smaller ones.

Usage Protocol:

• Priority on parallelized jobs is always given to users working on theses. This cannot be stressed enough.
• Absolutely no browsing of clearly questionable web pages. This also cannot be stressed enough.
• Abuse of the cluster will lead to job termination and locking of the account.
• Machines with console users should never be used for parallelized jobs without asking the console user first.
• Routine jobs (crons) should always be approved before use.
• For security, you can only ssh or sftp into the cluster through huey or metlab.
• Accounts with weak passwords will be locked.
• Your password should be different from other accounts you have.
• If you have a question on how to do something, or whether you can do something, ASK!
• The only public RAID5 partitions on the cluster are:
  • /home
  • /opt
  • /moe/GRIB
  • /moe/r0
  • /barney/r0
  • /maggie/r0
  • /frink/r0
  • /itchy/r0
  • /otto/r0
  • /scorpio/r0
  All other partitions are not redundant and have a much higher chance of data loss. Thus, back up your own files if you are using partitions other than those listed above.

Datasets:

• Real-time grids: Past 5 days.
  Locations: /moe/GRIB, /itchy/r0/GRIB2
• TC Track: Unisys: 1888-current
  Locations: /moe/r0/tracks
• TC Track: ATCF: 1888-current
  Locations: /moe/r0/atcf-archive
• CMC Operational Analysis: 2003-current [primarily TC season prior to 2004].
  Location: /itchy/r0/operational_analyses
• GFS Operational Analysis: 1998-current [primarily TC season prior to 2004].
  Location: /itchy/r0/operational_analyses
• NGP Operational Analysis: 1997-current [primarily TC season prior to 2004].
  Location: /itchy/r0/operational_analyses
• UKM Operational Analysis: 1999-current [primarily TC season prior to 2004].
  Location: /itchy/r0/operational_analyses
• CMC, GFS, NGP, UKM, GFDL, HWRF Forecast Grids: TC season only 1998-current.
  Location: On tape.
• NCAR/NCEP Reanalysis: 1948-current.
  Location: /barney/r0/reanalysis/ncep
• NCAR/NCEP Reanalysis2: 1979-2005
  Location: /barney/r0/reanalysis/ncep2
• ERA40 Reanalysis2: 1957-2002
  Location: /barney/r0/reanalysis/era
• NARR Reanalysis2: 1979-current
  Location: /barney/r0/reanalysis/narr
• 3B42: 1998-current
  Location: /itchy/r0/3B42
• SST: 1985-current
  Location: /itchy/r0/sst-nodc

Ownership (does not include implied sysadmin by Hart):

• Hart: 53%
• Chassignet: 20%
• Reasor: 10%
• Erlebacher: 5%
• Zou: 3%
• Ellingson: 3%
• Ruscher: 2%
• Maue: 1%
• Winterbottom:1%