NDMP backup performance degradation

[vkravindran]

Hi,

Since we migrated our TSM environment to a new server hardware, we are experiencing slow performance with NDMP backups.

Old environment

SUNFire v480 server
Solaris 8
LP10000 HBA Cards
LPFC driver 6.02f
DMX-3 backend disks
EMC Celerra datamover (DART 5.5.32)
TSM server 5.4.2.0
NDMP v4
IBM Ultrium 2 (LTO2) fibre drives (in Quantum P7000 library)

New Environment

SUNFire T2000 server
Solaris 10
LP10000 / LP11002 HBA Cards
LPFC driver 6.02 f
DMX-3 backend disks
EMC Celerra datamover (DART 5.5.32)
TSM server 5.4.2.0
NDMP v4
HP (LTO3) fibre drives (in Quantum PX720 library)


As you can see, the only things that changed between the old and new are the server hardware, addition of a new model of HBA card, and the library/drive. The same set of SAN disks (for TSM) were imported to the new server (Veritas diskgroup).

But we are seeing a drastic performance degradation in NDMP backups. NDMP Backups that used to go ~20-30 GB / hour now goes ~8GB / hour.

PS: We are also seeing poor backup performance to diskpool, after this migration.

Any help / advise to fix this is greatly appreciated-

Not able to get much help from EMC / SUN and IBM.

Thanks
R