iSCSI performance for Veeam Repository Target

[remko.de.koning]

Hi guys,

Not sure if my post is suitable for this forum but I thought I would give it a try nonetheless.
We have bought a Synology RS2416RP+ which we use as a backup repository for Veeam (Backup Copy Jobs)
I am not sure if I am getting the performance I need and I was hoping some of you have experience with this subject.

Below is the post I posted on the Synology forum but I doubt this forum is very active.

We have bought a Synology RS2416RP+ and have added 6 X 6 TB western digital red nas
In a RAID-5 config this results in approx 27 TB of disk space.
This diskspace is added as a ISCSI lun.
The Synology has 4 NIC's. 1 is used for management and the other 3 are used for iSCSI.
Each NIC has its own little network/subnet with a Server 2012 R2. No switch in between.
No extra configuration has been done on the NIC's (no experience on this subject)
MPIO is installed on Server 2012 and three iSCSI targets have been mapped.
The disk is available in Windows and is used for extra backup storage.
When monitoring in DSM I see the 3 NIC utilitzed evenly. (approx 50 mb/s)

Question:
I expected the performance to be a bit faster than the performance I am seeing now.
According to the performance monitor in DSM I have between the 140 and 180 Mb/sec on throughput.
The disk utilitsation is about 25%.
Would it be possible to gain more throughput or is the most I can get with these kind of disk.
I realize these are not enterprise grade disk on which the troughput is much faster.
What kind of IOPS could I expect?

thanks for any advice you might have.

Remko

[final]

I have noted a very significant performance difference when the iSCSI target LUN on the Synology is "file level" on a btrfs volume (using a "simple" LUN, not an "advanced" lun). I'm talking about >100% difference in performance compared to block level LUNs.

With a Raid6 over 12 6TB WD Red Pros, I get > 1GB/s, so the bottleneck is the 10gbps link in my case.

Also, for measuring throughput, I recommend Atto Disk Bench. Make sure you use a large test file (4 - 8GB, so that you can rule out cache). The interessting block sizes are >64k.

I get my 1GB/s at a block size of about 1MB/s (which is not uncommon for the backup job).

[remko.de.koning]

Wow. That is definitely a lot more than I have. Unfortunately I have already copied over several TB's to the new volume.
It is not that easy moving around large amounts of data like this. I will try to test your recommendation with a couple of smaller disk that I have around somewhere.
Thanks for the feedback.
I read somewhere that you can play around with block sizes. Although I have no idea how to do this I was wondering if this would make any significant difference.