I'm surprised to hear people having issues with a 3584.
I've got TSM servers with 3494's and 3584's; the 3584 is my favorite
TSM library, hands down.
Robotics are FAR superior to the 3494 - I can tell you bizarre stories
about problems with 3494 robotics.
The 3494 is slow and clunky with a lot of wasted motion compared to the
3584, which is fast and smooth. Inventory time is 30 SECONDS PER FRAME,
In that respect I certainly don't consider it a "step backwards".
For new TSM admins, the 3584 is much easier to deal with than category
codes (which can get ugly if you've ever had tapes misplaced due to
manual mode mounting). I've never had problems with a 3584 (or any
other SCSI library, for that matter) "losing" its inventory on a regular
basis; if you do I recommend raising a HIGH PRIORITY issue with hardware
support and get it addressed - SHOULDN"T HAPPEN on an Enterprise class
library.
Yes, I know that people are "used to" the way the 3494 works, but
functionally there aren't that many differences, depending on how you
set it up.
The 3584 web interface makes partitioning easy as pie - just a few
clicks. Then the library itself can manage the partitioning based on
VOLSER ranges, WITHOUT category codes - I think that's a step FORWARD!
I really like all the stuff you can do, easily, through the 3584 web
interface.
With the ALMS software and the VIO option, the 3584 I/O door "queues"
input and output just like the 3494 does (i.e., if you try to eject more
than 16 cartridges, the robot waits until you unload the I/O door
instead of failing the eject) (The only thing weird is that you still
have to specify SEARCH=BULK when inserting carts with TSM, even though
the cartridges have been moved from the I/O door to internal slots, they
are still "virtual" I/O slots.)
You can get dual grippers & HA for the 3584. It supports 3592 & LTO,
which is nice for some customers.
You can also get FICON controllers now (they are in a separate rack
called a 3953) which gives you a 3494 style console and connectivity to
a mainframe. In that case, the mainframe still uses SMS & category
codes, and it thinks its talking to a 3494. BUT, the Open Systems hosts
connected to it still see it as a SCSI library.
The BIGGEST FUNCTIONAL DIFFERENCE:
The 3584 doesn't have "manual mode" for TSM if your gripper is down.
(The 3953 does support manual mode for the mainframe component.)
If manual mode is important to you, you have to stick with the 3494.
Otherwise, the 3584 is totally slick.
Just my opinion....
Wanda
-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU] On Behalf Of
Steve Roder
Sent: Friday, February 02, 2007 12:47 PM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: Re: 3584 vs 3494
> This question is nominally directed to Richard Sims because it cites
> his Quickfacts. But, other knowledgeable people are welcome and
> encouraged to respond.
>
> In the '3584' entry in "ADSM/TSM Quickfacts" it is said that, "The
> 3584 is a "SCSI library": the host has to direct its actions. If
> you're accustomed to the 3494 and its automated actions, don't expect
> anything like that in the 3584. As one customer said: The 3584 is
> pretty much an automatic tape mounter with bulk storage and no real
> 'smarts'.
>
> In the 'SCSI Library' entry it is said that, "The TSM server must
> physically manage its actions, and must keep track of volume
> locations. ...the work they don't do they shift to the host, and so
> TSM is burdened with a lot of intricate SCSI element details and
> control issues."
>
> Do these entries describe current 3584 automation features,
> especially vis-a-vis TSM? If they do, how will TSM server
> performance and administration change for the worse if we migrate
> from 3494 technology to a 3584?
We have both a 3494 and a 3584 in our dual-server TSM configuration,
soon
to be 4 servers.
Our 3494 is an HA config with dual grippers on each robot. For this
reason, all of our primary data is in the 3494, and the 3584 contains
the
copy, which is SAN attached about 5 miles away. If our 4 server config
uses a 3584 for primary data, unlike the 3494, it will have single
points
of failure that the 3494 does not have.
To me, this is a step backwards. Is there an HA option for the TS3500
yet that provides dual library managers and dual robotics along with a
manual mode of operation? I guess if it did it would be a scsi-3494,
and
the beginnings of an acceptable replacement for a 3494 with HA frames.
I much perfer the 3494's categories to scsi elements. Sharing is simple
vs. partitioning.
Steve Roder
University at Buffalo
(spr AT buffalo DOT edu | (716)645-3564)
|
