|
Logical Volume Manager from a Technical User Perspective: Disk Management for
Data-Intensive Computing Research
Eva Hocks
In 1999 I came to San Diego from Germany to work at the San Diego Supercomputing
Center (SDSC). SDSC, an organized research unit of UC San Diego, develops and
applies advanced computing technologies to demanding computational science problems.
In doing so, SDSC is leading the effort by the National Partnership for Advanced
Computational Infrastructure (NPACI) to deploy a national metacomputing infrastructure
and foster scientific research.
I am the lead system administrator for the ASCI Blue Horizon IBM SP system.
The name Blue Horizon heralds the dawn of a new age in scientific discovery
and reflects SDSCs collaboration with IBM. Blue Horizon is the most powerful
computer available to the U.S. academic community and currently the 8th most
powerful computer in the world. Recent upgrades boosted the machines peak
performance to 1.7 teraflops 1.7 trillion operations per second and added
a higher-speed switch that connects the systems 152 nodes.
The SDSC archival file-storage systems support storage requirements ranging
from a few hundred GB to several terabytes (TB). The SDSC High Performance Storage
System provides 1 TB of disk cache and up to 180 TB of storage in three StorageTek
silos controlled by a 23-node IBM SP system. The archive currently holds more
than 80 TB of data in 7 million files.
One of the three supercomputers that SDSC supports for the national research
community is an IBM RS/6000 SP with an AIX operating system. The IBM RS/6000
SP has 1052 375-MHz processors, each with 4 gigabytes (GB) of memory.
The RS/6000 SP AIX operating system has a helpful system management interface
tool (SMIT), as well as very useful debugging support with sysdumps, crash,
snap, and for the SP, the css.s
|