My server is 2.1.5.18 running on an RS-6000 (SSA disks).  I, too, have
millions of objects in my ADSM database.  When complaining about
performance on queries or restores that use wildcards, I'm told that I
should spread my database over multiple physical disks.  Supposedly ADSM
will perform parallel queries.  I'm willing to restructure my database
if this will make a significant improvement but I can't seem to get
confirmation.

The single biggest caveat to this notion is that ADSM apparently doesn't
"spread" the database evenly over the multiple disks as it grows.  Is
there a way that this can be forced?

Tom Fluker
Viasystems Technologies

> -----Original Message-----
> From: Purdon, James [SMTP:james_purdon AT MERCK DOT COM]
> Sent: Friday, June 05, 1998 9:31 AM
> To:   ADSM-L AT VM.MARIST DOT EDU
> Subject:      Re: Anyone else having server database contention
> problems?
>
> Hi,
> We have observed this behavior (on server 2.1.0.13) when the client
> hadd 4
> million files contaned within a single file space.  However, it
> ususally
> took two simultaneous wild card operations for the client to hang the
> server.
>
> Our opinion is that large filespaces need to be divided into virtual
> mountpoints in order for ADSM to function effectively.
>
> Our database is about 26.5 Gb and contains entries for approximately 6
> million objects.
>
> Jim
> > ----------
> > From:         Bill
> Quintrell[SMTP:Bill_Quintrell AT PROVIDENTCOMPANIES DOT COM]
> > Reply To:     ADSM: Dist Stor Manager
> > Sent:         Thursday, June 04, 1998 6:01 PM
> > To:   ADSM-L AT VM.MARIST DOT EDU
> > Subject:      Anyone else having server database contention
> problems?
> >
> > Bill Quintrell@PROVIDENT LIFE
> > 06/04/98 06:01 PM
> > I don't know where along the way it started happening (we have
> millions of
> > objects in each ADSM server)...
> >
> > Simply put, if a client issues a wildcard request against the ADSM
> server,
> > then other client requests are effectively locked for a significant
> amount
> > of time.  I am not talking about moving any data - just the server
> > searching its database.
> >
> > We have tested this on OS/390 and AIX servers with similar results.
> Its
> > almost like the second and succeeding clients which need to store an
> > object
> > (insert a record into the database) do not get a fair shot at the
> server
> > until the first "query" has completed.   As an example, start a
> restore of
> > "xxxx*" - which does not exist.  While it is running, do a store of
> a
> > single, small file.  It may take over 20 times longer than normal
> and with
> > only 2 clients and virtually no data to transmit.
> >
> > Maybe there is a server tuning knob based upon the client or type of
> > server
> > request...
> >
> > Have you seen this?  It makes me wonder - what is the practical
> limit on
> > the ADSM database size for a given environment...
> > (Since we use ADSM as part of our online imaging system, one query
> can
> > start a hundred users screaming...)
> > Not to start a contest (our record is 52 hours for one incremental
> > backup),
> > what are some of the larger ADSM server databases - i.e. number of
> objects
> > managed?
> >