Hello,

I see no answer to this on the list. So get one
TSM utilizes "transactions", in other words it groups files until they
reach some limit and processes the whole bunch at once. Limits are size of
transaction in  kB (controlled by TXNBytelimit option in UNIX dsm.sys or
Windows dsm.opt) and number of files (controlled by TXNGroupmax option in
dsmserv.opt).
Every transaction will go into one volume and next transaction will go
into next volume in round-robin. If you have more client sessions (be it
many nodes or few nodes with many threads) than volumes at least two will
go to same volume. For volumes of type DISK this is not problem, that why
they are random-access volumes. Requests are performed immediately by TSM
and are queued by OS device driver (and may even be reordered by SCSI
disks or RAID controllers).
For sequential volumes (FILE or TAPE) things are as you are afraid - if
ten nodes try to backup to tape but you have only five drives, five
sessions will wait for resources until one of the other five session
completes.
If you have large disk you can define single volume of type DISK. It would
handle simultaneously multiple sessions even when is the only volume in a
storage pool. Creation of several volumes over same spindle would not help
and may even decrease the performance due to TSM trying to write
parallelly which would force HDD head to go back'n'forth. You will
improvement of transfer times (microseconds) at the price of greater seek
time (milliseconds) !!
If you create a filesystem and define stgpool of type FILE over this large
HDD then define maxcapacity to be smaller, maxscratch to be higher and
higher maxnummp to avoid congestion for mount points (each file serves as
virtual tape). So for FILE class you should ensure load to be spread more.
Hope this helps.

Zlatko Krastev
IT Consultant




Please respond to jdschn AT attglobal DOT net
Sent by:        "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>
To:     ADSM-L AT VM.MARIST DOT EDU
cc:

Subject:        Use of TSM disk storage pools volumes by clients

Greetings,
    Sorry if this is a newbie question, but I have not been able to find
the answer in the manuals.
    Say a TSM server has a disk storage pool with 5 volumes.  When a
given client is backing up to a disk storage pool, will all his backup
data go to the same disk storage pool volume?  Or will it go to one
volume only until that volume fills up, then go on to another volume?
    And what if there are 10 clients backing up to 5 volumes?  Do 5
clients get exclusive access to the 5 available volumes (like they would
to tape) while the other 5 wait, or do they all get to write to the
volumes?
    Another question that depends on the answers to that question is, if
you have a big disk, say a 73GB disk you intend to use for disk storage
pool, is it better to make it one big disk storage pool volume, or make
it more small storage pool volumes, like 7 volumes of 10GB each?  If
each client got a different disk storage pool volume, then you would
think this would perform very slowly, since you would have to seek all
over the disk as those 7 clients each wrote to separate volumes on the
same physical disk.  On the other hand, if you made it one 73GB volume,
it might perform very slowly if only one client could write to it, and
the others had to wait.

Please reply via email if you have any insight into these questions, and
I will post a summary to the list.

Thanks in advance,

John Schneider
Lowery Systems, Inc.