Skip to main content
Start a Conversation

Unsolved

This post is more than 5 years old

newcustomer1

3 Posts

365

October 6th, 2015 07:00

DD4500 - TSM - stgpool backup to LTO5 slow

Hi

We are a relatively new DD4500 user.  Our DD4500 is running as a VTL and is connected to our AIX TSM server via 8Gb FC SAN.

There are currently two fibre ports on the TSM Server zoned to 2 ports on the DD4500 with the access group so that there are 6 virtual tape drives on each fibre port.

The LTO5 drives are spread over seperate FC adapters on theTSM server.

If I do a stgpool backup with 6 processes reading from the DD4500 and writing to 6 LTO5 tape drives I am only getting around 350Mb/s , peaking at around 400Mb/s occasionally.  This is while there are no other client backups or restores taking place, no other TSM Server processes running and no cleaning running on the DD4500.  The TSM server has plenty of CPU and memory.

If I do a similar thing with 6 stgpool backup processes going from LTO3 tape to LTO5 tape I get much better performance, at least 750Mb/s

So I'm not sure where to start finding the bottlneck.  I'm really hoping the DD4500 can push out data to tape faster than this otherwise I have a real problem.

Any help or pointers appreciated.

Thanks

dynamox

9 Legend

 • 

20.4K Posts

October 8th, 2015 08:00

i don't know what resources you have available but could you test backing up DD4500 pool to a pool class disk (will need to present some SAN storage). I have not messed with TSM much since version 6.3.3

jbrooksuk

208 Posts

October 9th, 2015 06:00

There are quite a few things that need to be checked and it's not very easy to be conclusive on what you need to look at but I'll give you a list of where I might start.

  • Are the physical tape drives and virtual tape drives zoned/using the same 2x HBA host initiators?
  • How many shelves does your DD have?
  • Are you cleaning more than once a week?
  • How full is your DD generally?
  • Are there ISL's between your VTL and your physical tape?
  • How old/long has the backup been sat there before you cloned it?
  • Is your dedupe ratio quite good on what you are cloning from VTL?
  • Have you changed the block size you use on either VTL or physical tape?
  • Are you using single target/single initiator zoning methods?
  • The initiator (host) is definitely running 8Gb FC
  • The Target (DD) is definitely running 8Gb FC
  • Did you stipulate only one Primary FC endpoint in the access group and 'no' secondary (not 'all').

It should go faster but all of the above will make a difference.

regards, Jonathan

Was this post helpful?

View All

No Events found!

Top