On Mon, 27 Jan 1997, Bill Nickless wrote:
> I moved to ADSM after working with early versions of NSL Unitree. I
> apparently misunderstood the meaning, under ADSM, of the 'Next Storage Pool'
> option on a storage pool to include staging (from the next storage pool) as
> well as migration (going to the next storage pool.)
>
> Imagine a configuration where one has the clients access a disk storage pool
> which points to a tape storage pool for migration. As I now understand it,
> one could configure the disk storage pool to buffer tape writes, but once
> the data goes to tape I don't see how you can configure ADSM to store
> recently-recalled data to a disk buffer in case it will be accessed again.
>
> This has dramatic implications on system design; in order to buffer recent
> tape reads (taking advantage of temporal clustering of read access) one must
> use something like the HSM client.
>
> Is my current understanding correct, or have I missed something?
I think your analysis is correct. To mimic UniTree functionality, you
need the ADSM HSM client; neither ADSM Backup nor Archive are replacements
for UniTree. ADSM Archive (with its builtin expiry of data and its
node-oriented name space) has apparently been designed with short-term
archive in mind.
We have recently migrated 500 GB of UniTree data to an ADSM HSM system
(IBM 7015 R40, SSA disks, 3494 robot with 3590 drives). So far our users
are very satisfied with the improvements in performance and accessibility
of data (all standard UNIX file transfer protocols, use of graphical ftp
clients), and our operational staff appreciates the system's stability
and ease of operation. ADSM's HSM has not fully matured, yet, and
there is room for improvements (e.g. 2GB file size limit, performance
issues, management of data, etc.).
> How does one submit formal requests for enhancements to the ADSM product?
ADSM's HSM functionality might be what you are looking for.
Backstaging of Backup/Archive data to disk should be a configurable
option, based on the ratio of a client's restore performance vs. the
ADSM server's tape performance (throughput, positioning). With a fast
client connected via a fast network, you typically do not want to have
data staged back to disk in a restore operation. If a client is slow,
however, you want to buffer data on disk to prevent the client from
monopolizing a tape drive for hours.
Sincerely, M. Fink
--
Dr. Michael Fink +-----------------------------+------------------------
Dr. Michael Fink +-----------------------------+------------------------
EDV-Zentrum | Universitaet Innsbruck | Tel: +43(512)507-2311
Computing Services | Technikerstrasse 13 | FAX: +43(512)507-2944
--------------------+ A - 6020 Innsbruck, Austria | Michael.Fink AT uibk.ac DOT
at
=========================================================================
|
