iSCSI benchmarking is difficult due to all the various factors that play into it, just like NAS. I'm a UNIX only user, but I'd recommend the 3 best tools out there: 1) IOzone: http://iozone.org/ 2) Bonnie++: http://www.coker.com.au/bonnie++/ 3) PostMark: http://www.netapp.com/tech_library/3022.html

Bonnie++ is the best of the real quick and easy benchmarks around. IOZone is the best way to see whats really happening and where the sweet spot is, but more time consuming and confusing for the uninitiated. And PostMark is NetApp's own benchmark which does some interesting tests better representing the real world.

If your using a software initiator make sure to carefully examine CPU and interupt activity on your system during the benchmarking session.

benr.

PS: Here is an example of Bonnie++. I ran this from my Linux desktop system (with X and 20 terms, etc running). I'm using JFS on an iSCSI LUN provided by an F840 with a single shelf of 36G drives. I'm using a Linux software initiator and using the public switched 100M ethernet that I use for everything else. This test used a 2G sample. Here's the numbers: [benr@nexus benchmark]$ bonnie++ -d /na_iscsi/bench_scratch/ -s 2048M -m "F840 iSCSI" (Lines removed) Version 1.03 ------Sequential Output------ --Sequential Input- --Random- -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP F840 iSCSI 2G 3802 57 11296 12 5330 5 4322 69 10489 5 290.0 2 ------Sequential Create------ --------Random Create-------- -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete-- files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP 16 3842 32 +++++ +++ 3120 21 2000 35 +++++ +++ 1986 24

-----Original Message----- From: owner-toasters@mathworks.com on behalf of Bui, Marcus Sent: Tue 9/28/2004 1:08 PM To: toasters@mathworks.com Cc: Subject: iscsi benchmark tools Hi Toasters,

I just setup a iSCSI test environment and comparing the performance with a server's local disk. The target is the R200 (50GB LUN) with a W2K initiator, MS's iSCSI driver on 100MB NIC (Interl Ethernet NIC not iSCSI HBA). I used two different freeware tools (Roadkil Speed Disk v1.1 and IOMeter v200.7.30) and get totally different results.

iSCSI volume Random Read:

Roadkil - 30.3MB/sec

IOMeter - .06MB/sec

iSCSI volume Write

IOMeter - .03MB/sec

Roadkil - N/A on freeware version

Based on the Random Read numbers, it seems the two results are vastly different. Don't know which one depicts the 'true' performance for the iSCSI volume. Can someone recommend a tool you use before to benchmark iSCIS performance?

Cheers,

Marcus Bui AIM Investments IT Services phone: 713-986-9061 fax: 713-821-9524 email: marcus.bui@aiminvestments.com pager: 8773450839@archwireless.net

----------------------------------------- Confidentiality Note: The information contained in this message, and any attachments, may contain confidential and/or privileged material. It is intended solely for the person or entity to which it is addressed. Any review, retransmission, dissemination, or taking of any action in reliance upon this information by persons or entities other than the intended recipient is prohibited. If you received this in error, please contact the sender and delete the material from any computer.