Backup throughput dropping off dramatically

There is a utility called RawRead which measures the throughput of harddrives. This is to help find out if the problem is at the drive end.
http://www.drivesnapshot.de/freedos/rawread.htm

Another thing to do is enable logging on the switch during a backup. A lot of times this is an easy way to see on which side the problem is. For example  the port going to the server has little to know errors but the port going to the library has lots. At that point the problem has been cut in half.

I have not used EMC stuff but it probably has it's own logging that can be enabled.

To test to see if the bottleneck is at the tape drive end run the command "simulate -ca". This lets you create an imaginary tape device. So data is read from the source and is processed in the normal way up to the point of being written to tape where it is just dumped since there is not a physical drive. In this case it might be a pain to test. Any BEB Distributed servers will have to have BrightStor shut down and only the Primary running. Simulate will replace the current configuration so after removing the simulated device it will be necessary to run Device Confguration again to setup the library. That's why it is a bit of a pain.

That brings up a good point, if there are Distributed servers check to see if they have the same problem.

Does the BEB server have just one HBA?

Actually the Brightstor Server that is having the issue is a distributed server with one HBA that connects to the SAN and another HBA that connects to to the Library.