Tyler,
I don't have any statistics to look at from your site, and you didn't say if
you weere observing any"latency" when the data was flowing in chunks.
Your subset request is a much smaller amount of data than the WMO .* pattern,
so it is likly that you would observe more continuous data flow with a full
pattern. Note that the WMO pattern includes both IDS|DDPLUS as well as HDS, and
there is certainly a lot of of HDS data (eg models) that your pattern below
doesn't capture.
What others have already mentioned in response is that you definitely want to
ensure that your feed requests to different upstream hosts are either
identical (to take advantage of autoshifting), or completely disjoint.
If you are observing latencies, pleease provide that information.
Steve Chiswell
Unidata User Support
On Mon, 2 Apr 2007, Tyler Allison wrote:
> [for ldm-6.6.0 and ldm-6.6.2]
>
> If I use the following feed pattern [upstream = idd@wisc and bigbird@tamu]:
>
> request WMO
> "((WF|WO|WU|WR|WW|WG|AC|NW)US)|^S[AP]|^S[NS]|^WT(PZ|PN|PA|NT)|(^S[IMN]V[^GINS])|(^S[IMN]W[^KZ])|(^S(HV|HXX|S[^X]))|(^SX(VD|V.50|US(2[0-3]|08|40|82|86)))|(^Y[HO]XX84)|^A
> CUS11|/p(FTM|FWDDY)" upstream.server.edu
>
> the data seems to flow down to me in "chunks". What I mean by this is that
> is when I run 'ldmadmin watch' I see no data for between 4 and 10 minutes.
> Then all of a sudden the last 4-10 minutes worth of data comes down within
> a 3 second period. It then holds again...and 4-10 minutes later another
> payload of data comes down...etc...etc.
>
> If the *ONLY* thing I change is the feed pattern to:
>
> request WMO ".*" upstream.server.edu
>
> The data feed runs normal.
>
> I'm confused! I, and the upstream, have bandwidth to burn so I'm leaving
> the feed as ".*"...my pattern was attempting to conserve bandwidth for the
> upstream providers but it seems I did more harm than good.
>
> I havent totally verified this issue but it seems to be the culprit. Can
> anybody confirm?
>
> -Tyler
>
>