Bandwidth on the SAN

[David Devlin_1]

The read time from our disks on the san is terrible. We use netbackup 4.5 datacentre for our backups. All of our servers boot from the san as well as having all of their data located on the san too. All servers use gigabit adapters. All of our backups run between 1.5mb per/sec and 2mb per.sec. This should be a lot higher. The servers are under no stress while the backups running, I have tested a backup to a null device so as to eliminate the tape libraries but the throughput is still the same. Do you think it maybe something to do with the switches or just that the san cannot handle the amount of data transfer from all the servers?

Regards David

[Uwe Zessin]

David,

if you have brocade switches you can try the 'portPerfShow' command to watch. A data rate of 1.5 to 2 MegaBytes per second is nothing to a fibre channel switch. In a test environment with 1 GigaBit switches I was able to get about 80 MegaBytes per second even over an inter-switch-link.

I am not sure I have understood your setup correctly. You read from the disks through a fibre channel interconnect and send it over GigaBit ethernet to the backup server?

Have you heard about the 'StorageWorks Library and Tape Tools (L&TT)', yet? It allows you to do a read test of your data on the SAN. Here is the URL:
http://h18000.www1.hp.com/products/storageworks/ltt/index.html

Let us know how much it could get from your array.

[BR651180]

Hi David
It seems that from you query the problem is most likely in the storage front or the resouces may not be sufficient.
Which storage box you are using- XP or EMC, if either of the above please check the Host adapters (whether 2Gbps port) and SAN switches (this also should support 2Gbps) and the HBA (also 2Gbps) then there should not be an issue.Please check IOs on the storage (if EMC use tool EMC WLA)so based on that you may need to increase the HBAs on the host side or additional host port on the storage side.
Please do check that are you crossing the recomended capacity per host port on the storage.
On the server front please check the IO/sec and from the application front check the IO waits.
As you are using network backup please check the gig E port on the switch side, please also try scheduling the backups in the different time frames (not in the peak time of user).

[David Devlin_1]

Thanks guys, We are using 2x Compaq EMA12000's with dual hsg80 controllers on both. One is attached to 2gb switches and the other is attatched to 1gb switches. The servers that seem to suffer most are the one's that are connected to the 1gb switches, however, when checking these ports none of them are going beyond 10-11 mb per/sec in fact when monitoring all ports either switch never goes above 25-30 mb per/sec. This leads me to believe that the disks themselves may be under a lot of stress as there are 5 exchange servers, two file and print servers and a mixture of other servers all using the san disks for data and boot. I have done some performance monitoring on one of the busier exchange servers and no matter what disk performance counter you select it races up to an average of 200-300% time while the switch port that the server is plugged into only states 4-5mb per/sec. Do you think we have overloaded the san?

Regards
David