Performance is unique to each iSCSI SAN setup (i.e., switch manufacture, Host NIC/HBA, Blade Server setup, Array Firmware, number of members in a PS Array group, etc.). So there isn’t really any one “standard” measurement for performance goals. The best way to determine this is to setup the iSCSI SAN according to our published documents, run SANHQ (EqualLogic’s Performance monitor) for a week or longer, and then review the data to see where improvement (if any) can be made.
Regarding setup up a High performance iSCSI network, you can look at this link, and in the list, view “Dell EqualLogic Network Performance Guidelines v4.0”. Additionally, in the list you will also see another links to VMware setup, Performance, IOmeter, etc.
http://www.delltechcenter.com/page/Guides.
I'd like to be a bit more specific.
Looking at equallogic's white paper, you can use direct connected VM. It looks so simple, there is almost noting to do.
We decided however to compare the native ISCSI from VMWARE with IOMETER on Windows 2008 R2, and Windows 2003 SP2( 64 ko chunk, 100 % sequential. Just to compare) with the ISCSI from Microsoft (direct attached) and with RDM.
There is almost no difference between "the native ISCSI from VMWARE" and the RDM, but when you go for a direct attached VM, I got a fourth of the perf (200 M, instead of 800+) ! . It's not only performing worse, it sucks 60 to 90 % of my CPU !!
What about updating this famous "multi vendor post to help our mutual iscsi customers using VMWARE"
In 2004, we were booting on SAN from Clariion (DELL). In 2009 we migrated and booted from equallogic. But getting the HIT on VM's with decent performance seems out of reach.
Joe S586
7 Technologist
•
729 Posts
0
July 20th, 2011 07:00
Performance is unique to each iSCSI SAN setup (i.e., switch manufacture, Host NIC/HBA, Blade Server setup, Array Firmware, number of members in a PS Array group, etc.). So there isn’t really any one “standard” measurement for performance goals. The best way to determine this is to setup the iSCSI SAN according to our published documents, run SANHQ (EqualLogic’s Performance monitor) for a week or longer, and then review the data to see where improvement (if any) can be made.
Regarding setup up a High performance iSCSI network, you can look at this link, and in the list, view “Dell EqualLogic Network Performance Guidelines v4.0”. Additionally, in the list you will also see another links to VMware setup, Performance, IOmeter, etc. http://www.delltechcenter.com/page/Guides.
Also check out this link from our Storage Infrastructure team: http://www.delltechcenter.com/page/Storage+Infrastructure+and+Solutions+Team+Publications
For specific setup information see: support.dell.com/equallogic website.
Regarding VMDQ, not sure if you reviewed this: http://www.vmware.com/pdf/10GigE_performance.pdf
Regarding MEM, yes you could see significant performance gains (when properly setup).
Please post any additional questions you might have.
Regards,
Joe
PAQUOT
2 Posts
0
August 1st, 2011 15:00
Looking at equallogic's white paper, you can use direct connected VM. It looks so simple, there is almost noting to do.
We decided however to compare the native ISCSI from VMWARE with IOMETER on Windows 2008 R2, and Windows 2003 SP2( 64 ko chunk, 100 % sequential. Just to compare) with the ISCSI from Microsoft (direct attached) and with RDM.
There is almost no difference between "the native ISCSI from VMWARE" and the RDM, but when you go for a direct attached VM, I got a fourth of the perf (200 M, instead of 800+) ! . It's not only performing worse, it sucks 60 to 90 % of my CPU !!
What about updating this famous "multi vendor post to help our mutual iscsi customers using VMWARE"
In 2004, we were booting on SAN from Clariion (DELL). In 2009 we migrated and booted from equallogic. But getting the HIT on VM's with decent performance seems out of reach.