We have an L700 (690 slot) with 8 x DLT-7000 drives here.
It's being driven by a single Sun E-450: 4 x 480 Mhz CPU, 4 Gig Mem,
five PCI SCSI cards (one tape drive on _each_ bus, plus one bus for
the library), and a Gigabit-Fiber card. This server has done well for
past three years, even with all eight drives spinning.
Software is Solaris 2.6 and Sun/Legato SBU 6.1.3.
We use four 36 Gig internal drives because two of them are mirrors:
The boot disk is mirrored via dd, and the dedicated /nsr disk will
mirror on Solstice Disksuite (in progress anyway). A second disk bus
was added for redundancy.
The real issue for us now is the capacity of DLT-IV media. Hopefully
we'll have a chance to get better drives on-line next year (using
AlphaStor for migration access). That's when a storage node will be
brought on-line too, since we're admittedly near the edge a bit
already on I/O.
Hope this helps,
--TSK
===== =====
Timothy S. Kimball
System Administrator, 3rd/Night Shift tkimball AT brass DOT com
===== #include <std_disclaimers.h> =====
On Tue, 2 Dec 2003, George Sinclair wrote:
> Our network is a standard switched 100 mbit ethernet for most hosts and
> gigabit ethernet for a few of the heavy hitters. The backups run across
> the same wires as all the rest of the network traffic. Backups typically
> run at night so there's little network traffic to interfere. Compression
> is done on the devices, not the clients. Our original justification for
> the storage node was to free up some of the burden on the main server as
> more and more clients were coming under the backups, especially ones
> with large file systems. I guess we could have just bought a bigger,
> more powerful server, but the idea of the storage node, the cheaper cost
> of licensing, etc. jsut appealed to us.
>
> Also, having a storage node gave us the opportunity to play with running
> NetWorker under Linux without having to switch the primary server from
> Sun to Linux until we could first evaluate how well the libraries worked
> under Linux. If we'd never had these problems, I would have migrated
> everything to the Linux storage node, set it up as the primary server
> and then made the sun a storage node, keeping in mind, of course,
> possible problems with little versus big endian. Having a storage node
> server also seemed like a good way to increase the parallelism, but
> perhaps most importantly, it seemed like a good way to cover ourselves
> in the event that the library on the main server was not responding. I
> think having a storage node server is a nice thing and not so much a
> burden as long as it works, but we've had nothing but problems with the
> SCSI communication so this is really the justification for replacing it
> with a Sun to see how the SCSI works there.
>
> George
>
[snip]
STATEMENT OF CONFIDENTIALITY
The information contained in this electronic message and any attachments
to this message are intended for the exclusive use of the addressee(s)
and may contain confidential or privileged information. If you are not
the intended recipient, please notify SunGard Trading Systems immediately
at (201) 499-5900 and destroy all copies of this message and any
attachments.
--
Note: To sign off this list, send a "signoff networker" command via email
to listserv AT listmail.temple DOT edu or visit the list's Web site at
http://listmail.temple.edu/archives/networker.html where you can
also view and post messages to the list.
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
|
