ADSM-L

Re: Multiple ADSM Servers

1997-01-08 10:14:33
Subject: Re: Multiple ADSM Servers
From: Brian Pennington <brian_pennington AT MAILLINK.CMIC DOT COM>
Date: Wed, 8 Jan 1997 10:14:33 EST
     We are running multiple servers and sharing a 3494 with 4 3590s and
     2 3490e cart drives.  We have had 4 3590's running on one scsi
     controller with no problems.  We dont do this all the time but
     if one server needs a couple extra drives we set them up on the
     same controller.  Our servers are a sp wide node and a 58h.

     hope this helps

     Brian

     Brian_Pennington AT aici DOT com


______________________________ Reply Separator _________________________________
Subject: Multiple ADSM Servers
Author:  "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU> at SMTP
Date:    1/8/97 6:54 AM


Has anybody ever run multiple AIX ADSM servers all working with one 3494?
We're running the server on an SP wide node and have managed to shoehorn 4
SCSI controllers in for the 3590 drives in the library. We've still got
more client machines to migrate to ADSM and some of them have large volumes
of data to backup so 2 more 3590s have arrived. Problem is we know from
past experience that its just not worth putting more than one 3590 on a
SCSI controller.

The only answer we can come up with is to run a second ADSM server on
another node with adequate MCA slots free for the SCSI connections to, say,
the two new 3590s in the library and start putting clients onto this one.
This will bring more administration overheads for us and some possible
areas of confusion. Can anyone see any other problem areas running these
two ADSM servers against a single 3494, containing 6 3590 drives, two of
which are managed by one server, four by the other?

We're doing what we can to minimise tape drive contention but contention
does exist and hence the need for more drives. Anyone any experience of
running ADSM on a high-node? These have 16 MCA slots and would allow us to
continue with one ADSM server but we've heard that the high-speed switch
code doesn't perform well on an SMP machine.

Thanks in advance
Alan White
<Prev in Thread] Current Thread [Next in Thread>