This archive contains answers to questions sent to Unidata support through mid-2025. Note that the archive is no longer being updated. We provide the archive for reference; many of the answers presented here remain technically correct, even if somewhat outdated. For the most up-to-date information on the use of NSF Unidata software and data services, please consult the Software Documentation first.
Jim, I recall your saying you were running your Unidata software on freebsd. We just acquired a freebsd box too for our Unidata software using freebsd 4.7 release. Tom recommended I check in with you to see if you had any advice on Softupdates. Our box has 3 scsi drives and I've made 2 slices on the second, a 3 GB for the ldm queues, the rest for the Unidata software home(s), and on the 3rd drive I've got a slice for swap, the rest for ldm's /data. My question is: is it advantageous or no for the queue partition on the second drive to have Softupdates turned off, while all other partitions (except /) have Softupdates implemented? My limited understanding of Softupdates is that it improves I/O rates by 'cache'ing file metadata updates to disk, so that the ondisk metadata changes are allowed to 'get behind' corresponding data writes at times until I/O activity will allow them to catch up. Does this argue for or against Softupdates on the queue file system? On Mon, 2003-03-17 at 19:49, Unidata Support wrote: > >From: "Neil R. Smith" <address@hidden> > >Organization: TAMU > >Keywords: 200303130136.h2D1aoB2022568 LDM FreeBSD > > Hi Neil, > > >I've got my ldm.pq on a separate 3GB partition on a different disk from > >my /data partition .. thinking that would make things a little faster. > >I'm using Softupdates on my filesystems, but thinking about this queue > >partition, do you think I should not enable Softupdates for it? > > If the only thing you have in the file system is your LDM queue, I > would not have Softupdates turned on for it. > > >What config/partitioning have you found optimal on freebsd? (I'm using > >freebsd 4.7 release). > > Since we are not (yet) doing anything but moving data with our FreeBSD > box (we are using it in LDM stress testing and it is sailing along > happily), it is hard to say if the queue would be better in its own > partition. It is not on a separate partition on our system. For > reference, we are using 10K rpm IBM SCSI disks in our FreeBSD box. It > is also a dual 1 Ghz processor system with 4 GB of RAM. We run our LDM > with a 2 GB queue, but that is mainly since we are using it in testing > movement of all IDD data to multiple (7) downstream machines. FYI, it > is moving an average of 31 Mbps of data to downstream machines, and has > been doing so for just over 6 days now. After we stop our stress > testing series, we will be converting this FreeBSD box into our data > ingest/decode machine for GEMPAK, McIDAS, and the netCDF decoders and > as an ADDE server for THREDDS testing. > > Sorry I couldn't provide more insight into disk tuning. We may have > more to say after the machines takes up decoding and data serving > duties. You might want to touch base with Jim Koermer of Plymouth > State <address@hidden> to see what he did before he put a > RAID on one of his FreeBSD systems. > > Tom > **************************************************************************** < > Unidata User Support UCAR Unidata Program < > (303)497-8643 P.O. Box 3000 < > address@hidden Boulder, CO 80307 < > ---------------------------------------------------------------------------- < > Unidata WWW Service http://www.unidata.ucar.edu/ < > **************************************************************************** < > -- Neil R. Smith, Comp. Sys. Mngr. address@hidden Dept. Atmospheric Sci., Texas A&M Univ. 979/845-6272 FAX:979/862-4466