I would have to say that the performance of ADSM is very dependent on your exact
environment. In an environment with large databases and large file sizes, ADSM
works well. Performance is very good in these types of environments and the
small number of nodes to manage make ADSM a good solution.

In an environment with millions of small(<100K) files, ADSM starts to break
down. It has an extremely high database overhead that limits overall
performance. In our environment, we expect to peak out at about 1500kB/s, barely
acceptable performance. We are currently planning on running a large scale trial
through the end of the year and have been trying to put a lot of pressure on IBM
to fix the bottlenecks. Most of the performance enhancements are not scheduled
to happen until next year though. The only disappointment support-wise, is the
glacier like rate of development and bug-fixing. There are tremendous time lags
in making the latest server versions available on different platforms. Also,
support for new OS releases on client platforms seems to lag almost 6 months
behind the release of the OS, HP/UX 10.0 is a good example of this.

We have already run one small trial with 20 nodes and about 2 million files, and
are large trial will cover about 25 nodes with probably 5-10 million files.
Total data for the large trial should be about 500G.


Paul Shields
pshields AT bnr DOT ca

In message "ADSM bugs and support.", mmk AT isc.upenn DOT edu writes:

>Hello,
>
>
>I would like to ask ADSM users if many are happy with a product or support.
>
>A few month ago we installed ADSM on our SP2 machine, but we can't call
>it backup solution yet.
>
>Very often we have tape drive problems, sometimes these errors not in errpt
>but on activity log as OP=WRITE or READ with sense=00.00.00.00.
>These errors terminate process, after 20 GB of data already processed.
>
>Many features which in ADSM books or README files don't work, and I am
>getting fallacious answer, for example, DSMSCHED.LOG pruning will happen
>only after scheduler will be restarted, even IBM suggested to include SCHLOGR
>(days),[s n] in DSM.SYS. So why bother with this parameters, if our goal
>to do backup with centralize scheduler and automate backup as much as
>possible.
>
>Sometimes ADSM create duplicate names in filespaces and refusing to
>backup filesystem from clients nodes because of it.
>
>When we made marketing research of backup product, many IBM representatives
>claimed that in 4q 1994 we will be able to backup ORACLE online, now nobody
>even mentioning if they continue working on this interface, just referring
>to third companies.
>
>I have eggs on my face for recommending implementing ADSM.
>
>My questions are how to correct these problems, to whom to call to get
>quick response and resolutions, often we are waiting week to talk with
>level 2 and other week to get results.
>How people backing up 150 GB of ORACLE data base, knowing it have reliable
>backup in case of restore.
>
>IBM people, welcome to reply.
>
>Thanks.
>Mark
>
>
>
>
>                         --------------------------
>                        |   Mark M. Klivansky      |
>                        |University of Pennsylvania|
>            ------------|       E-MAIL :           |------------
>            \ voice:    |   mmk AT isc.upenn DOT edu      | FAX:      /
>             \ (215)    |                          | (215)    /
>             /898-1466   -------------------------- 898-0386  \
>            /                 |              |                 \
>            ------------------                ------------------
>