4 minutes ago, 1812 said:

 

I'm seriously beginning to think the MD1000 is the "issue." Thankfully I'm not as bad as this guy was but wonder if I could incorporate anything from his tweaks into my usage:

 

https://thias.marmotte.net/2008/01/dell-perc5e-and-md1000-performance-tweaks/

 

Ahh yes I have a poweredge c1000 with a PERC5/E and it has awful throughput. Raid 10 on that thing was still crazy slow if the reads were not contiguous.

I have an unsubstantiated idea after reading a bunch of info on the md1000.

 

I'm running it in split mode, 7 disks one side, 8 disks the other, each side goes to a different server in my use scenario.

 

What if, in split mode, it splits the total bandwidth to each side of the box? So you get the full x4 bandwidth in unified, but only x2 in split. Is such a thing possible?I have no way to actually test this because unRaid locks up if I give it the unified enclosure (for some unknown reason.) And I don't know of anyway to test if the connection to the box is x2 or x4.

 

Am I crazy?

11 hours ago, 1812 said:

I have an unsubstantiated idea after reading a bunch of info on the md1000.

 

I'm running it in split mode, 7 disks one side, 8 disks the other, each side goes to a different server in my use scenario.

 

What if, in split mode, it splits the total bandwidth to each side of the box? So you get the full x4 bandwidth in unified, but only x2 in split. Is such a thing possible?I have no way to actually test this because unRaid locks up if I give it the unified enclosure (for some unknown reason.) And I don't know of anyway to test if the connection to the box is x2 or x4.

 

Am I crazy?

 

It's a possibility, and the numbers would make sense.

 

Unfortunately I only know how to check the link between the HBA and the 1st expander, not to a cascaded expander.

 

Also since I had my expanders out for an upgrade I tested if there was any performance penalty due to cascading and for me there was, about 10%, HP expander single linked delivers 1000MB/s max when cascaded tn an Intel expander, not cascaded I get 1100MB/s.

 

The best way to test would be to connect the MD1000 directly to the HBA, you could test for improvements and also check the link width with the cat /sys/class/sas_host/ command.