20010727: accessing data via ADDE versus NFS

>From: Wayne Bresky <address@hidden>
>Organization: Cornell
>Keywords: 200107261641.f6QGft109510 McIDAS f2c display

Wayne,

>I have a very fundamental question as far as accessing data from a
>single machine in a network.  In your opinion, is it better to set up
>the machine as an ADDE server and serve the data to the rest of the
>machines or is it better to mount up the data from the single machine
>via NFS and allow the other machines access to it that way?

Since all data access in McIDAS is through ADDE, and since setting up
one central machine as an ADDE server greatly simplifies maintenance, I
recommend ADDE.  ADDE and NFS are not mutually exclusive, however.  You
can do NFS mounts for both McIDAS and GEMPAK use and still do ADDE
requests using ADDE.  This might seem like overkill until you
understand that ADDE transfers are typically faster than reading data
from NFS mounted disk files.  The reason for this is that the ADDE
"pipe" is bigger than the NFS "pipe".

>I'm leaning towards using NFS instead of ADDE because this method will
>work for GEMPAK as well as McIDAS.  What are your thoughts on this?

Do both.  The ADDE access on the server will require that you:

o setup the ADDE remote server on the server
o setup clients (users using McIDAS) to request those data from the
  ADDE remote server

What kind of machine are you going to be using for the server (NFS and
ADDE)?  The reason I ask is that there is a known problem serving
sounding data through the remote ADDE server on Linux systems running
the 2.1 or later kernels (we talked about this previously).  This means
that if your server is going to be a Linux box, you _will_ want to have
NFS access to the upper air data at least.  The ADDE setup for the
RTPTSRC dataset will then be LOCAL-DATA and each user's session will
have to be configured to know where to find the data files that make up
RTPTSRC.  We can go over this in detail when you are ready.  It is very
easy to do.

Tom