Data Throughput measurement - how to?

[Haggisnneeps2]

We are backing up over 4 TB data but backup performance seems unusually slow considering we have Gigabit from SAN to backup servers but the NICs during peak backup time seem to be doing next to nothing. I would like to find out if theres a bottleneck during the read operation from the MSA1000 or anywhere in between, looking at each individual component along the way ie physical disk read to cache, cache to buffer, buffer to NIC, NIC to SAN Switch, San switch to core switch, core switch to edge switch, edge switch to backup server NIC, NIC to buffer, buffer to cache, cache to physical disk (and anything in between i've missed). Along the way it would be nice to know if each device/component is achieving its theoretical/actual transfer speeds/specifications.Is there any way to do this? It seems unusual for us to be able to buy 4GB capable Fibre-SCSI routers but have Ultra320 speed RAID controllers or MSA20s that can't deliver anywhere near that kind of throughput. Thanks

[meredith shaebanyan]

Could you post more information about the source disk configuration? You mentioned an MSA 20 -- that is a SATA enclosure and the SATA drives provide considerably less performance than the Ultra320 drives, even though the MSA20 has an Ultra320 host-bus interface.

[Haggisnneeps2]

The source is an MSA1000 with 14x 300GB 10k drives attached to a DL380G4 via Fibre which in turn is into a SAN switch and then into a Cisco 3750G via fibre
The backup destination is a DL380G4 with a 6400 PCI-X Controller with 2x MSA20s attached each with 12x 750GB SATA. This in turn is connected via 4GB HBA to MSL6030 tape unit with 2x HP LTO-4 Ultrium 1840 tape drives (via 4GB Fibre to SCSI adapter HP AD577A)

[Haggisnneeps2]

Yes thanks - i understand that (space constraints meant we had to go for SATA for the backup drives)

What i need to know or be able to find out or even measure (somehow) is where in the chain the bottleneck is occurring and if its possible to do anything about it - ie is there a more effective way of partitioning the MSA1000 or a more effective array configuration strategy.
I'm about to introduce 2 new MSA30 drive trays to the MSA1000 and i'd like to make sure they are at absolute optimum performance in every respect

[meloni]

Ltt have some instrument that will help you to identify the source of the problem.

you have in the program a sytem performance test and a device performance test that allow this.

This link explain you how to proceed:

http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?objectID=lpg50460