Target Bottleneck?

[trichert]

I have a Veeam 6 backup server which backups up to an iSCSI target over 10GB direct connect (no switch inbetween) jumbo frames enabled on both ends. I can copy data between the 2 at about 350MB (big B) /sec.

When I run my SQL backup job it says

Bottleneck : Source 4% , Proxy 2%, Target 99% (network 0%) and it is only processing at 7MB/s

However, I can kick off my file server backup and while it also says target 99% it processes at 130MB/s

What am I missing? How is the Target the bottleneck if a second job can be kicked off and get 20x the bandwidth ?

[tsightler]

I'm assuming these are incremental runs? Are you using reverse incremental? Pay attention to the amount of data transferred as reported by the job, my guess is that the SQL server is having to read and transfer much more data than the file server. In most cases the issue is simply that the target has find performance for streaming I/O, but reverse incremental backups are not streaming I/O, they are random I/O, having to write new blocks and move old blocks to the rollback. In this scenario IOPS are far more important than streaming write throughput.