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

20020927: UFRJ as a primary IDD redistribution point for Brazil (cont.)



>From:  David Garrana Coelho <address@hidden>
>Organization:  UFRJ
>Keywords:  200209271806.g8RI6v115886 IDD Brazil Meteoforum

Hi David,

re: UFRJ as a primary IDD redistribution node for Brazil

>       We are very pleased to being considered to play such a role in
>your system, and more than willing to help improve the system to fellow
>institutions.

This is wonderful news, thank you!

>We currently run LDM/GEMPAK on a Intel P4/1.8GHz with 256Mb RAM,
>and 35Gb of disk. This machine also runs MM5 Model about 60% of time,
>daily. Currently around 8Gb of disk space is free. I believe you have how
>to check bandwidth capacity/velocity from where you are.

The requirements to be a relay will most likely force you to increase
the size of your LDM queue.  What size queue are you currently using?
It is most likely that you will need to increase the queue to at
least 500 MB and perhaps significantly more as more sites start
to feed from you.

What operating system are you running (e.g., RedHat Linux 7.1, Sun
Solaris x86 5.8, etc.)?

>I am attaching my current ldmd.conf and pqact.conf files, so you
>can see our current configuration.

Thank you for including these files.

We will want you to do several things in your LDM setup to better
handle data ingestion and relay (I will provide details in the
coming days):

1) upgrade to the current release of the LDM, LDM 5.2.1.  This release
   is only available in source form, so you will be forced to build
   the new distribution from source.  You can get the LDM 5.2.1 distribution
   in source form by anonymous FTP from our FTP server, ftp.unidata.ucar.edu:

   <login as 'ldm'>
   cd ~ldm
   ftp ftp.unidata.ucar.edu
      <user> anonymous
      <pass> address@hidden
      cd pub/ldm5
      binary
      get ldm-5.2.1.tar.Z
      quit

2) change the data requests in your ldmd.conf to:

   o split the IDS+DDPLUS feed into thirds (better insures that the
     products get their with little to no latency)

   o add ingest of the Unidata-Wisconsin (LDM 5.1.x MCIDAS or LDM 5.2+
     UNIWISC feedtype).  The current GOES-East (GOES-8) images in this
     datastream are useful for northern Brazil.  We will be adding
     sectors tailored to Brazil in the coming weeks.

   o add/modify ingestion of global model data (the non-global model
     data available in our HDS stream is useful for the continental
     US, Canada, Mexico, and the Caribbean only).

   o change the host from which you are receiving data (we will switch
     you to a machine we maintain at the National Science Foundation
     in Washington, D.C.

   o add the reporting of Realtime Statistics (available with LDM 5.2+)
     in addition to, and eventually replacement of 'pqbinstats'.

   o we may be asking you to upgrade your LDM somewhat frequently
     as we are focusing a lot of effort on improving its design
     especially for long distance data delivery (like to South America).

Again, I will provide details on the ldmd.conf changes in the coming
days.  I can tell you that our recommendations are a direct result
of what we learned in sending data to UFPA.

Now, what you do with the data when it is received is up to you.  The
existence of the produts in the LDM queue is sufficient to relay them
to other institutions.

>We want you to know that if along this
>testing we find there are greater requirements of processing power, disk
>space or bandwidth, we can work this around, according to this needs and
>our possibilities here.

This is very good news indeed, and very generous of you!

If you are running Linux, it would be advantageous to add more memory
to your machine (actually, adding more memory can never hurt :-).

It is possible that when you start looking at the other products that
are flowing in the IDD, you will want to start filing them to disk for
your own use.  This would most likely mean that you will need
additional disk resources to store and use those products.  I don't
know what prices are like in Brazil, but in the US, hard disk and RAM
prices for PCs are so low that augmenting these resources costs next to
nothing (for instance, 256 MB of RAM is something like $30 and 40GB of
disk is on the order of $60).  I run my home system with 640 MB of RAM,
would have added more except that I need to upgrade the CPU from its
current 400 Mhz to something more modern, like 2 GHz.

We can proceed with the establishment of UFRJ as a principle IDD relay
in a couple of different ways.

1) Ricardo was kind enough to give me logins on his machine
   (desmata.ufpa.br) so that I could install new versions of the LDM as
   soon as they become available (they also gave me 'root' access so I
   could do things like install NTP and set the setuid bit on LDM
   programs like rpc.ldmd and hupsyslog).  This is what allowed me to
   quickly learn how the LDM could to be tuned to ensure reliable,
   timely delivery of data to Belem.  The amount of tuning that will be
   needed at UFRJ should be less since the network connection to it is
   so much faster than to UFPA.

   One of the other reasons that I would appreciate a login on your
   system is so that I can run network tests from your machine to those
   sites that will want to feed from you.  Since the running of these
   tests is sporadic, it is hard to predict and schedule needed access
   times.

   Giving me access to your machine would mean that the initial
   setup would be done by someone else, so you would not have to worry
   about fitting the tasks into your schedule.  The downside, of course,
   is that you would be giving access to your resources to someone you
   have never met.

2) The other approach would be for me to relay the information on the need
   to upgrade the LDM installation on your machine as needed.  This will
   not be a problem especially after reliable data delivery has been setup.

>       Best regards,

I look forward to working with you and your UFRJ collegues on the
establishment of what I hope will be a service used throughout Brazil!

>David Garrana
>Researcher
>Laboratorio de Prognosticos em Mesoescala - LPM (www.lpm.meteoro.ufrj.br)
>Departamento de Meteorologia - IGEO/CCMN/UFRJ

Cheers,

Tom
--
+-----------------------------------------------------------------------------+
* Tom Yoksas                                             UCAR Unidata Program *
* (303) 497-8642 (last resort)                                  P.O. Box 3000 *
* address@hidden                                   Boulder, CO 80307 *
* Unidata WWW Service                             http://www.unidata.ucar.edu/*
+-----------------------------------------------------------------------------+