[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

20020304: hosting ADDE access to Unidata NEXRAD composite imagery (cont.)



>From: Eirh-Yu Hsie <address@hidden>
>Organization: Aeronomy Laboratory/NOAA/DOC
>Keywords: 200203041624.g24GOMK11535 McIDAS ADDE NEXRAD composite

Hsie,

>It is fine for you to set up the new decoder in cumulus.al.noaa.gov.

OK, thanks.  I will try to get this done today.

>The disk space is not a problem.  Disk I/O is.
>You did not tell me how often the data will come in (every 15 minutes?).

Sorry, I should have included the data frequency.

Expanded relevant size information is in order:

Product                    File size after decoding  Frequency
--------------------------+-------------------------+--------------
10 km national RCM           166880 bytes            30      min
1 km regional floater N0R    361280 bytes            7 - 10  min
6 km national N0R            401280 bytes            7 - 10  min
1 km national N0R          14208533 bytes            10 - 15 min

>Right now I am running decoder on a Sun Sparc 3000 (4X167 MHz CPU with
>1.5 GB memory).

Uncompressing/decoding (convert from PNG to AREA or GINI depending on the
product) does not take much CPU.

>The data are stored in a A1000 array (Hardware RAID 0).

This is the same kind of unit that is attached to motherlode.  Given
our experiences with modtherlode, I am suprised that disk I/O would be
a concern of yours.

>I am planning to move decoders to a PC-Linux machine. Do unidata has any
>opinion or suggestion?

Our experience with Linux, gathered by working with user sites that are
using Linux on their LDM machines, suggests that a Linux box running
the LDM should have more RAM than the exact same box running Solaris
x86.  We do not have experience with running RAIDs under Linux, but
we do have experience under Solaris x86.  We have had very good results
running both software and hardware RAID under x86.

We were just about to contact sites running Linux to see if they would
be interested in participating in an experiment using the Web100 tuning
techniques developed in part at NCAR.   In order to participate, a site
would need to take the absolute latest Linux kernel (a lot newer than
the ones that come with Linux bundles from RedHat, Debian, Slackware,
etc.) and build a custom version with the Web100 stuff built in.
Web100 mods are aimed at adjusting networking parameters on the fly
based on usage and connectivity metrics.  You can learn more about
Web100 at the Web100 home page:

http://www.web100.org/

Please let me know if you are at all interested in participating in this
test.

Tom