[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Platforms #QXN-439339]: New hardware suggestions?
- Subject: [Platforms #QXN-439339]: New hardware suggestions?
- Date: Thu, 19 Oct 2006 15:21:43 -0600
Hi Pete,
re:
> Any idea on whether the ldm benefits much from threading? I.E. is it
> worth pursuing dual core CPUs or is it more just best to have lots of
> RAM?
This is an interesting question that is not easy to answer simply:
We convinced ourselves that having lots (e.g., 4) of CPUs was not useful
on a system that both ingesting data from several (exact number is hard
to pin down) upstreams and relaying data to lots (> 100) downstreams.
Our view was created from performance testing on our old toplevel IDD
relay node, thelma.ucar.edu, which was a SunFire 480R.
At the same time, we are now testing two SunFire x4200 machines (dual
dual core Opteron 2.2 Ghz w/16 GB RAM) in our IDD cluster. This test
is very new, so I don't have much to report yet (other than the units
appear to be working nicely).
> Right now I'm spec'ing out systems with dual single-core opterons,
> 2Ghz for the ingest/relay and 2.6 or 2.8 for the decoder. 8 Gb or 16 Gb
> on the relay, and 8 Gb on my decoding box.
Hmm... so the question of the hour is if a single dual core Opteron is
better or worse LDMwise than a box with dual single core Opterons. I
wish I knew the answer to this one!
> It's looking like I can get something like this and around 5 Tb of
> external (scsi) SATAII raid all in a rack for around or a little more
> than my 16K.
Sweet!
> I'm thinking 8 Gb of RAM should be adequate for us for both the
> ingest/relay box and our decoder/server box. 8Gb should hold at
> least an hours worth of all feeds we're getting.
An 8 GB LDM queue will hold between 5300 and 12000 seconds of data
for all of the feeds relayed by our cluster:
CONDUIT
DIFAX
FNEXRAD
FSL2
GEM
HDS
IDS|DDPLUS
NEXRAD2
NGRID
NIMAGE
NLDN
NNEXRAD
UNIWISC
> Thanks
I'm not sure the info above is what you are looking for, but it is the best
I can do. One of our "problems" is we are continuously changing equipment
in order to learn the answers to the kind of questions that you are asking
_before_ users ask them. Lots of fun :-)
Cheers,
Tom
****************************************************************************
Unidata User Support UCAR Unidata Program
(303) 497-8642 P.O. Box 3000
address@hidden Boulder, CO 80307
----------------------------------------------------------------------------
Unidata HomePage http://www.unidata.ucar.edu
****************************************************************************
Ticket Details
===================
Ticket ID: QXN-439339
Department: Support Platforms
Priority: Normal
Status: Closed