Hi,

until November 2005 everything in my TSM world looked fine, until we needed
a restore of our exchange db:

We were not able to restore (had been tested in the past) what we needed.
The restore always ended with failures regarding the data on the tsm server.

I invoked IBM and we found, that the data at the TSM Server were bad. We
were told to use CRC checking for our nodes.
Doing CRC checking we found, that a lot of our "successful" backups where
indeed bad!
Especially the tdp agents that send large objects have problems with crc's
since retries are time-consuming and you get easy new crc's.

So we are looking now with a lot of efforts into our network environment,
but I wonder
- why it is necessary to use crc in TSM despite there is TCP/IP that should
take care of that errors
- how crc checking is been done within tsm (e.g. it seems for each full
received backup object, differences data/protocol checking, number retries,
...)

(CRC checksums for the storagepools are not used at the moment, there is no
indication for that)

Does anyone has information's or experience's with crc checking?

Environment
TSM Server Windows 2003 5.3.1.2
TSM Clients 5.2.3.11 Windows 2000/2003
Backup to Disk Clariion CX700 with ATA Drives
GB Ethernet Backbone (4 Switches Nortel)