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

Re: Data latencies: could much of this be avoided? (fwd)


===============================================================================
Robb Kambic                                Unidata Program Center
Software Engineer III                      Univ. Corp for Atmospheric Research
address@hidden             WWW: http://www.unidata.ucar.edu/
===============================================================================

---------- Forwarded message ----------
Date: Mon, 13 Dec 1999 18:56:14 -0500
From: Peter Schmid <address@hidden>
To: C. Vandersip <address@hidden>
     address@hidden
Subject: Re: Data latencies: could much of this be avoided?

The reason for this is that if you seperate out the feeds into two requests
with one of the being the hostname  and the other being an IP then the ldm
issues two seperate pqact's for the two processes.  So any backup in one feed
does not effect the other feed.

Pete.

Peter Schmid
Sr. Programmer/Analyst SUNY at Albany
Department of Earth and Atmospheric Sciences
Phone:(518)-442-4571
E-Mail:address@hidden


C. Vandersip wrote:

> Interesting note Gilbert.  Can anyone provide some concrete answer to
> this?  We currently have our feed request set up using the following
> format:
>
> e.g.,
>       request WMO|FSL2       ".*"    yang.sprl.umich.edu
>
> The LDM Config Checklist from the Workshop has the format as:
>
>         request UNIDATA  ".*"  bar.bar.edu
>
> Would it be better to split into:
>
>         request IDD|DDPLUS ".*"  bar.bar.edu
>         request HDS        ".*"  anothersite.edu
>
> Are you're results still holding, Gilbert?
>
> Regards,
>
> Chris
>          ###############################################################
>          #                      Chris Vandersip                        #
>          #        Computer Research Specialist/Dept. Sysadmin          #
>          #  Rm. 024, Dept. of Meteorology, Florida State University    #
>          #          address@hidden   (850)644-2522                #
>          ###############################################################
>
> On Fri, 10 Dec 1999, Gilbert Sebenste wrote:
>
> > Hello all,
> >
> > One of the things that was a hot topic in the UNIDATA Newsletter this fall
> > was the worry of the data feed latencies via the LDM.
> >
> > While I know they are going to go a different route in the future sometime
> > to handle faster feeds, I'm wondering if there isn't something we can do
> > right now to stop a lot of the problem we're seeing.
> >
> > Pete Pokrandt reminded me during the NOAAPORT circuit board fry to split
> > the NP feed into two when you request it from a site, channeling the data
> > into two pipelines instead of one. He does it, and he doesn't get
> > latencies (he's inside the same building, granted). That is, when you
> > request NOAAPORT data from a site, do this:
> >
> > request IDS|DDPLUS noaaport.serverwhatever.atmos.edu
> > request HDS 10.30.28.22
> >
> > I forgot all about this...and since I started doing it today, there have
> > been no problems. Granted, traffic is a tad lower today on our partial T3,
> > but not by very much. And, for the first time, we're not getting
> > latencies. We were only getting them in the afternoon, and not fatal;
> > usually around 30-45 minutes. But now, they've disappeared. That even with
> > a scrambled full NIDS feed coming through the pipeline on the SDUS5*
> > headers (downstream sites, I'll get rid of those shortly...sorry about
> > that). That means I know I'll be able to handle the data feed on October
> > 1...and I want ALL the NIDS data from ALL the sites. So I'll be sucking in
> > anything with SDUS5* on it.
> >
> > Just a suggestion for those who are having trouble. Maybe the latencies
> > are mainly LDM created, as opposed to net-congestion.
> >
> > And maybe I'm full of it. But I wonder how many of us are doing it, and
> > if everyone did it, would it help stop the latencies?
> >
> > *******************************************************************************
> > Gilbert Sebenste                                                     
> > ********
> > Internet: address@hidden    (My opinions only!)                     ******
> > Staff Meteorologist, Northern Illinois University                      ****
> > E-mail: address@hidden                                 ***
> > web: http://weather.admin.niu.edu                                      **
> > Work phone: 815-753-5492                                                *
> > *******************************************************************************
> >
> >
> >