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

20040217: LDM testing at SJSU -update (cont.)

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 Note that the archive is no longer being updated. We provide the archive for reference; many of the answers presented here remain technically correct, even if somewhat outdated.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.

  • Subject: 20040217: LDM testing at SJSU -update (cont.)
  • Date: Tue, 17 Feb 2004 09:44:57 -0700
>From: Mike Voss <address@hidden>
>Organization: SJSU
>Keywords: 200402010010.i110ACp1023488 IDD test

Mike,

>Thanks for the feedback. Late in the day yesterday I realized I did not
>have the CONDUIT feed on rossby split out into different requests,
>i.e:
>
>request CONDUIT   "[05]$" thelma.ucar.edu
>request CONDUIT   "[16]$" thelma.ucar.edu
>request CONDUIT   "[27]$" thelma.ucar.edu
>request CONDUIT   "[38]$" thelma.ucar.edu
>request CONDUIT   "[49]$" thelma.ucar.edu

When I wrote the reply this morning, I didn't realize that you had
added the CONDUIT feed to rossby (i.e., I hadn't look at the rtstats
pages).

>I found out how important that is.

Yes, splitting the full CONDUIT feed is important when the sending
LDM is not electronically "close" to the receiver.  Combining
all feeds with a:

request ANY     .*      upstream.feed.host

really can only be done when machines are on the same LAN.

>After I did split the feeds, rossby
>was able to catch up quickly as can be seen in the dramatic drop in
>latency yesterday afternoon (~22Z).

The drop in latency from 10000 seconds down to 30 seconds is
dramatic indeed!

>So, I'm a little hessitant to say this, but it appears our tests are
>complete. I can get the full feed into my office here with limited
>latency, in fact, I have two full feeds coming in now, rossby and
>methost24.

Two feeds of all data in the IDD into a domain with little to no
latency is good proof that there is no packet shaping going on.

>Once again, thanks so much for your help. By having multiple machine
>names allowed on emo, I was able to move my machine around campus and
>really prove to the network authorities that we have issues. It turns
>out that our Alcatel switches are really the problem, and not any
>traffic shaping.

Ah Ha!

>And even though I'm finally getting the full feed now,
>we are getting all new cisco switches and a 10 Gig backbone in the next
>few months. Hopefully it's clear sailing from here out.

It really should be clear sailing with a 10 Gb backbone.  Congrats!

Cheers,

Tom
--