Understanding Windows disk throughput

[robnicholsonmalt]

Idle question for a Sunday afternoon and not necessarily Veeam related. I run Veeam B&R on my home server. It's currently doing a full backup from a RAID-0 array with 2 x 2TB SATA-3 drives in a USB-3 external disk enclosure. The backup is writing to a 4 x 4TB SATA-3 RAID-0 array in an eSATA external disk enclosure. This is Resource Monitor:



How come it appears to be writing about twice as fast as it's reading? That doesn't make sense!

[Regnor]

You're reading the data from 2 drives and writing the data to 4 drives? In theory RAID0 gives you n performance, where n is the number of drives.
So for your case this would explain why the performance of 4 disks doubles compared to 2 disks.

I wouldn't go with RAID0 and it's 0 fault tolerance by the way.

Edit: After posting I did understand your issue. So you can ignore my reply
Edit2: What kind of job and processing mode is this?

[robnicholsonmalt]

This is my belt-n-braces backup drive. Copies of virtual machines and nothing that would cause any upset if it was lost. The main data drive is RAID-10.

It's a normal Veeam B&R recovery job - running on the same server as Veeam. So it's a server backup. I'll grab some more details when I get home.

[robnicholsonmalt]

Ahh, I think I've twigged this one. You were indeed correct in focusing on the number of disks in the RAID array. The figures given in Resource Monitor are not average speeds - they're more the maximum speed achieved. So the 2-disk striped source disk can handle two read operations at once whereas the 6-disk striped target disk can spread the writes across all 6 disks.

So the source disk is reading @ ~100MB/s but it's then writing that data twice as fast to the target disk. In effect the target disk will be idle for half the time whilst it waits for the slower source disk.

When one looks at the throughputs in Veeam itself, it's the speed of the slowest device - the source in this case.