I'm facing a phenomen, which I can't explain and understand - maybe somebody can help me ?!
Problem:
We're using DeDuplication for the firststage-backup, afterwards the data is copied to external media - this time USB-disks.
During the copy-job from the Dedup-folder onto USB-disks (one active) the speed is between 950 and 1250 MB/min - not very fast but ok.
The USB-Disks are attached via USB3 and only one concurrent job os allowed to be written onto the USB-disks.
Now the strange thing starts:
When a second job starts while another is running, the speed drops to 250-400 MB/min.
But the second job can't really start as long as the first one isn't finished (because only 1 concurrent write is allowed to the usb-disc), so imho it shouldn't take any performance from the system.
Why does the speed drops, when the second job doesn't really do anything (because the target-device is blocked)?
Environment:
BE2012 with actual updates. The hardware is a HP DL180G6, CPU: E5606, 12 GB RAM. 5 SAS 1 TB Harddisks attached to an HP SmartArray P410 with 256 MB Cache and battery. CPU usage below 50%, RAM-usage ~ 50%.
Comparison to normal B2D-job:
A copy-job from a "normal" B2D-device (located on the same disk-channel) runs with 2.000-2.500 MB/min (which means the disc and usb-channel isn't the bottleneck). So the Dedup-device takes about 50%performance for rehydrating the data - which is "as designed and ok.
Any idea ?
