Solved

Maximising Tape Throughput?

  • 20 December 2021
  • 3 replies
  • 434 views

Userlevel 1
Badge +5

I have an FC connected IBM LTO8 library writing LTO M8 media through Commvault.

MA is Windows and client is Windows.

On the subclient the application read size is 2028KB as the volume for that subclient is ReFS formatted.

What chunk and block size would people suggest on the data path to the library please?

It’s on the defaults right now.

I’m experimenting with data readers and the obvious things and keeping an eye on latency on the source storage but I’m trying to just get this data moving to tape as fast as is possible.

Thanks :grinning:

icon

Best answer by Damian Andre 21 December 2021, 13:48

View original

3 replies

Userlevel 7
Badge +23

Multiplexing factor is the obvious one. On the job can you identify the bottleneck - there should be an indicator specifying if its read, write or network.

Userlevel 1
Badge +5

Damian thanks, so on a job with 8 readers and multiple readers allowed in a mount point (multiplexed to a single stream/drive) I’m seeing 3% read and 96% network.

I presume as the library is a local FC drive on the same server as the storage (the server is an MA) this is just how Commvault reports local throughput as it’s definitely not exiting the server :grinning:

Userlevel 7
Badge +23

Hi @Paul Hutchings , hope all is well.  Following up on this thread.  Did you get what you needed via @Damian Andre ‘s reply or did you need more clarification?

Reply