Need advice with a storage server configuration

[will.daw]

I am needing some advice on a storage server I built for GFS retention. As of now, I have three RAID 10 servers that hold 30 restore points and I want to create a backup copy job, so I can move it to 14 restore points. Recently, I built a 36-bay Supermicro server and stuffed it with 36-4TB 7.2K RPM SAS drives with a LSI MegaRAID SAS 9271 controller card and configured the drives in a RAID 60 (3 groups) with a 64k stripe size. The server connects to the Veeam Windows 2012 server by iSCSI with a 64k NTFS unit size. I have the server local to the datacenter right now, but I will be moving it offsite when I get everything configured right. The networking behind it is a 10-gig fiber backbone, but I am using a 1 gig NIC because it will be a 1 gig connection offsite.

My backup copy job is 6.5TB of data and I am experiencing slow processing with our file servers, exchange, and oracle db servers. Other VMs seem to process fine, but it hits a halt on these servers and it never completes the first full backup in the allotted time. I did redo the job and changed it to continuously, and it finished 4 days later, but according to Veeam they are increments and not a full backup (this is seen in the restore window when you restore a VM from the backup copy job). I am seeking a solution to increase the processing time, so my question is should I increase the RAID stripe size? Or, break up my backup copy job? Or, both?

Summary of Backup Copy Job
Processing Rate: 17 MB/s
Bottleneck: Target
Load: Source 18% > Proxy 11% > Network 17% > Target 73%

I did discover yesterday that the RAID 10 servers have a 64k stripe size, but the NTFS is a 4k unit size on the Veeam Windows server. Could this be a hindering point as well, since my backup copy job is referencing the backups on those RAID 10 servers? The Veeam backup jobs are set to the default (Local Target). Any suggestions will be appreciated.

Also, thank you guys for the forum and blogs because it has been a great resource. I have finally hit a point where I need to reach out and ask for advice instead of guess work. Shout out to Luca’s blog http://www.virtualtothecore.com/en/veea ... ripe-size/ because I have read over and over.

[Gostev]

Yes, you want your RAID stripe size at 256KB at least, as typical Veeam I/O block size is around 512KB assuming all the default settings. The math is as follows: default block size we operate with is 1MB, and that gets compressed 2x on average.

NTFS cluster size does not affect performance, but you want it at maximum to reduce fragmentation (which does impact performance).

[will.daw]

Awesome! Sounds like I have a some reconfiguring to do. Thank you for the help and quick response. I always enjoy and look forward to "The Word from Gostev" each week, keep up the great work.