[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference cookie::archive_backup

Title:	Archive/Backup

Moderator:	COOKIE::MHUAIG

Created:	Wed Sep 08 1993
Last Modified:	Fri Jun 06 1997
Last Successful Update:	Fri Jun 06 1997
Number of topics:	479
Total number of notes:	2283

388.0. "Central ABS Catalog between server and client(s) using DFS" by CX3PST::WSC217::SWANK (David) Wed Feb 26 1997 09:43

\DAVID LAWERNCE with TELECHECK in HOUSTON is asking;

Is it functional possible to have the disk that the ABS catalogs are on be
DFS server from the ABS server cluster to other ABS client clusters and
standalone nodes where each system populates the same catalogs and the
ABS server can initial client restores?

\David

T.R Title User Personal
Name Date Lines

388.1 COOKIE::LEWIS Mon Mar 03 1997 13:16 35

T.R	Title	User	Personal Name	Date	Lines
388.1		COOKIE::LEWIS		`Mon Mar 03 1997 13:16`	35
	The short answer is: no, don't do that. It is "possible" to have the catalog on a disk that is DFS served. HOWEVER, you probably don't want to do this because it will be a LOT of data going over the network while a save is running. Depending on what is being saved, there may be a lot of records being added to the catalog file, and that requires sending all of that data over the net. This may not only be slow for the updates, but it may impede the performance of other network activities. This all depends on the speed and capacity of the network, but it in any case, it will be an impact. Right now the catalogs are normally stored on the "client" system, for VMS, and the "server" can initiate a restore operation for a client. For UNIX and NT clients, the catalog is stored on the VMS "server", but the server can still initiate restores for the client. In fact, UNIX and NT restores must be started on the VMS server. Finally, the NODE name is currently NOT used for restoring VMS files. That is because at the time we added the nodename, vms catalogs were assumed to be on the individual VMS client, and it was assumed there was a separate catalog for each VMS client. This is still the assumption for VMS clients. We needed to store nodenames for UNIX and NT clients because the server stores the catalogs for UNIX and NT nodes. SO, if you have multiple VMS client nodes writing to the same catalog (whether it is DFS served in a network, or even cluster nodes with different physical devices, but the same logical name, (it is called disk1 on both systems, but points to physically different devices), you would have to have separate catalogs, or we could not determine which "disk1" you wanted. We are looking at adding the nodename to VMS lookups also. code is there, we just need to test to be sure everything works. Hope this helps. Jim

The short answer is: no, don't do that.

It is "possible" to have the catalog on a disk that is DFS served.  HOWEVER,
you probably don't want to do this because it will be a LOT of data going over
the network while a save is running.  Depending on what is being saved, there
may be a lot of records being added to the catalog file, and that requires
sending all of that data over the net.  This may not only be slow for the
updates, but it may impede the performance of other network activities.
This all depends on the speed and capacity of the network, but it in any
case, it will be an impact.

Right now the catalogs are normally stored on the "client" system,  for VMS,
and the
"server" can initiate a restore operation for a client.  For UNIX and NT
clients, the catalog is stored on the VMS "server",
but the server can still initiate restores for the client.  In fact, UNIX and
NT restores must be started on the VMS server.

Finally, the NODE name is currently NOT used for restoring VMS files.  That is
because at the time we added the nodename, vms catalogs were assumed to be
on the individual VMS client, and it was assumed there was a separate catalog
for each VMS client.  This is still the assumption for VMS clients.
  We needed to store nodenames for UNIX and NT clients
because the server stores the catalogs for UNIX and NT nodes.  SO, if you have
multiple VMS client nodes writing to the same catalog (whether it is DFS 
served in a network, or even cluster nodes with different physical devices, but
the same logical name, (it is called disk1 on both systems, but points to 
physically different devices), you would have to have separate catalogs, or we 
could not determine which "disk1" you wanted.  

We are looking at adding the nodename to VMS lookups also.  code is there, 
we just need to test to be sure everything works.

Hope this helps.
Jim