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

20030224: LLDM 6 testing

>From: Carl Sinclair <address@hidden>
>Organization: CAPS
>Keywords: 200302242020.h1OKKvt03684 LDM6 testing

Hi Carl,

>So far, LDM 6.0.0.17 seems to be performing well with 3 radars. 

OK, we are currently receiving 55 radars with 6.0.0.14 on our toplevel
IDD relay machine, thelma.ucar.edu, with good performance.

>However, we are about to setup an in-house test that will allow us to do
>much more than we are able to currently.  Do you have a utility that you
>use to test LDM? Something that can generate a large volume of data or
>simulate a large volume of data?  I think this would really help us in
>our testing.

Steve Emmerson created an LDM6 prototype feeder that moved synthetic
200 KB products and compared throughput using it and LDM5 between two
machines located here at the UPC and in Belem, Brazil.  The test was
very specialized in that neither LDM was using a product queue. The
test was designed to measure the difference in moving data using
blocking RPCs (LDM5) and non-blocking RPCs (LDM6).  Steve is out sick
today, but I will ask him how hard it would be to make his test
platform available to you.

The other possibility is for us to feed you the CONDUIT stream and you
use it as a high volume stream in your test.  CONDUIT data (high
resolution NCEP model output) volumes vary between about 1.2 GB to 1.8
GB per hour in products that can be up to a couple/few hunderd KB in
size.  Would this help you out in your testing?

>Thanks for the help.

No worries.  Please let me know about the CONDUIT feed.  If you want
it, please give me the fully qualified name of the machine you would be
running your LDM on so I can allow the feed on a machine here in the
UPC or at the ATM offices of NSF in Washington, D.C.

Tom Yoksas