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

[Support #SBB-325304]: Re: 20110712: CONDUIT request -- fire weather grids

Hi Becky,

re:
> So let me ask you one more question then, while we're on the subject...
> the WOC also offered up more than 3 servers at each location.  I kind of
> thought that might be overkill since there's a limited number of top
> level systems pulling from the WOC.  But I really should ask you --
> would you see any benefit in having more than 3 LDM servers at each WOC
> location?

I just talked with our system administrator about this (Mike Schmidt; Mike
Jeff Weber, Steve Emmerson and I are the ones "responsible" for the IDD
setup/running here in Unidata), and we agreed on three things:

- there is really no need for more machines in the present setup

  Meaning that if the idea is to continue sending the same kinds of volume
  in CONDUIT to the same small set of relay nodes, then adding more
  machines will not make things better or worse.

- but, the more machines the merrier :-)

- it _would_ be useful to add more machines to each site _if_ the volume
  of data and number of downstreams was significantly increased

  Our approach would be to setup a cluster of machines modeled on the
  one that we run here in UCAR: an "director" (frontend) that forwards
  data feed REQUESTs to a the least loaded "real server" (backend)
  that is operating as part of the cluster.  The input into the
  "real servers" would be through one or more "accumulator" machines.

  This approach would allow you to service a more diverse community
  of users (the U.S. weather enterprise) directly.  This is sort of
  like what Louis Uccellini was probably driving at in his comment
  during Linda's talk, even though he was putting the onus on purchase
  of machines on the private sector.  In my mind, it shouldn't matter
  who actually purchases the machines to be used... privately purchased
  machines could function as "real servers" in a clustered data
  deliver approach as well as if they were purchased by NCEP.  There
  would, of course, have to be an agreement by the private sector
  purchasers that the machines could be used in the clustered approach
  and monies provided to do routine maintenance, periodic upgrades
  (like increased RAM, new disks, etc.), and the like.

  Just so you know, I have long wanted to see this kind of setup for
  real-time delivery of high volume model output by NCEP.  The FTP servers
  are great, and so is NOMADS (really great), but most users would probably
  want the data pushed to them in real-time so that they could initialize
  their models faster.
 
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: SBB-325304
Department: Support CONDUIT
Priority: Normal
Status: Closed