Thomymaster
Contributor
- Joined
- Apr 26, 2013
- Messages
- 142
Hi guys
I run my FreeNAS 8.3.1p2 for supplying my ESXi via ISCSI.
old CPU: Intel XEON 5060
new CPU: Intel XEON E5345
RAM: 8GB
HDDs: 4x Seagate Barracuda 7200.11 (ST31500341AS) configured as a RaidZ2 with "force 4K sector"
ZIL: none
L2ARC: 64GB SSD
When i set it up i noticed a very bad ISCSI 4k write performance and enabled the tunable (vfs.zfs.no_write_throttle) as seen in this thread:
http://support.freenas.org/ticket/1830
However when i had my old 2-core CPU installed in the FreeNAS host this didn't improve performance. Now that i have a quad-core CPU the tunable works, see the 2 below screenshots (CrystalDiskMark is installed in a Win7 x86 VM on my ESXi).
Is there a reson behind it why the results differ so much or is a quad-core CPU needed to make the tuneable work?
Cheers
Thomy
I run my FreeNAS 8.3.1p2 for supplying my ESXi via ISCSI.
old CPU: Intel XEON 5060
new CPU: Intel XEON E5345
RAM: 8GB
HDDs: 4x Seagate Barracuda 7200.11 (ST31500341AS) configured as a RaidZ2 with "force 4K sector"
ZIL: none
L2ARC: 64GB SSD
When i set it up i noticed a very bad ISCSI 4k write performance and enabled the tunable (vfs.zfs.no_write_throttle) as seen in this thread:
http://support.freenas.org/ticket/1830
However when i had my old 2-core CPU installed in the FreeNAS host this didn't improve performance. Now that i have a quad-core CPU the tunable works, see the 2 below screenshots (CrystalDiskMark is installed in a Win7 x86 VM on my ESXi).
Is there a reson behind it why the results differ so much or is a quad-core CPU needed to make the tuneable work?
Cheers
Thomy