Data Protection for SQL creates a separate TSM Server session
for each stripe and then waits for the SQL Server to send data
to each stripe. The SQL Server determines which data goes
to which stripe, and writes the data to it.

>From what you have put into this append, it appears as if
sending the data to the TSM Server is the bottleneck in
your scenario.

How many tapes are mounted for the log backup?
How many TSM Server sessions are being started for the log backup?
Can you examine the session statistics on the TSM Server?
What is happening on each session?
How much data is being sent for each session?
What is the speed like when you use a lower number of
stripes for the log backup? (For example, using 2 stripes.)
Finding the bottleneck on the TSM Server/Storage Agent is the
key to this one.

Thanks,

Del

----------------------------------------------------
"ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU> wrote on 09/20/2004
05:39:02 PM:

> TSM Client V5.3.2.1, TDP for SQL V5.2.1, TSM Storage Agent V5.2.3.2
> TSM Server V5.2.3.2.
> Windows 2003 Itanium Cluster with 2 servers.
>
> We are doing some SQL backup testing today.  We have a 2.5 TB
> database that we backed up at 477 GB/hr to 10 tape drives using
> LANFree (2.5 TB's in 5.1 hours).
>
> We started a logs backup of a 13 GB Log and it has been running for
> nearly 90 minutes so far.  The backup mounted a tape via LanFree
> right away and has had the same tape mounted in the same tape drive
> this whole time. There are no errors being reported anywhere.
>
> On the command window where we started the TDP backup we see
> "Waiting for TSM server" with pages of periods ".".  Then we see a
> message showing that is wrote some data.  It keeps repeating that
> process.  At the rate being displayed of 620.12 Kb/Sec it looks like
> the Log file backup is running at around 2 GB/hr.  Any ideas what
> might be going wrong?
>
> We still have all the settings maxed:
> Buffers 8
> Buffersize 8191
> SQLBuffers 0
> SQLBuffersize 4096
> Stripes 10
>
> These same settings didn't seem to effect the full backup adversly?
>