i'm running a similar setup to yours except i'm using iSCSI HBAs and my array is configured for Raid 50. your results are a little slower than mine. what kind of switches are you using? i'm assuming your are running gigabit?

i see you are running 3.0.2. just putting this out there...there is a large thread about slow response times over the SC network. wondering if this is impacting you because you're using the software initiator.

do you have a 3.0.1 box to try your test on?

here is the latency thread

http://www.vmware.com/community/thread.jspa?messageID=729853&tstart=0#729853

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
TABLE oF RESULTS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

SERVER TYPE: Physical running Windows 2003 SP2
CPU TYPE / NUMBER: 2 Intel DualCore 3.6GHz
HOST TYPE: IBM HS20 Blade, 8GB of RAM
STORAGE TYPE / DISK NUMBER / RAID LEVEL: Xiotech Magnitude 3D connected via Brocade 48k fabric switch and a Brocade switch embedded in the blade chassis.

##################################################################################
TEST NAME-------------------Av. Resp. Time ms------Av. IOs/sek-------Av. MB/sek------
##################################################################################

Max Throughput-100%Read........___11.6_______..........___4829_______.........____150______

RealLife-60%Rand-65%Read......_____35_____.........._____1144_____.........____8.9______

Max Throughput-50%Read.........._____27.7_____..........______1545____.........____48.3______

Random-8k-70%Read.................___32.3_______..........____1173______.........___9.2_______

EXCEPTIONS: CPU Util.-XX%;

##################################################################################
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
TABLE oF RESULTS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

SERVER TYPE: Virtual Machine running Windows 2003 SP2 on a VI3.0.2 Host
CPU TYPE / NUMBER: 2 AMD DualCore 2.4GHz
HOST TYPE: IBM LS20 Blade, 16GB of RAM
Guest Type: 1 VCPU, 4GB of RAM
STORAGE TYPE / DISK NUMBER / RAID LEVEL: Xiotech Magnitude 3D connected via Brocade 48k fabric switch and a Brocade switch embedded in the blade chassis.

##################################################################################
TEST NAME-------------------Av. Resp. Time ms------Av. IOs/sek-------Av. MB/sek------
##################################################################################

Max Throughput-100%Read........___15_______..........___3297_______.........____103______

RealLife-60%Rand-65%Read......_____25.2_____.........._____1919_____.........____15_____

Max Throughput-50%Read.........._____57_____..........______956____.........____29.9______

Random-8k-70%Read.................___20.5______..........____2306______.........___18_______

EXCEPTIONS: CPU Util.-XX%;

##################################################################################

Here are my first results

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
TABLE oF RESULTS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

SERVER TYPE: Physical W2003 x64 enterprise 16 Gb ram.
CPU TYPE / NUMBER: 8 cores pysical
HOST TYPE: HP DL 585, 16GB RAM; 4x Opteron Dual Core2.8 Ghz
STORAGE TYPE / DISK NUMBER / RAID LEVEL:SanMelody 10 G ram, 10x 146 G 10k SAS raid 1+0 - Write caching turned off on the physical SAS drives
Host adaptor FC QLA2460

##################################################################################
TEST NAME-------------------Av. Resp. Time ms------Av. IOs/sek-------Av. MB/sek------
##################################################################################

Max Throughput-100%Read........_5.2861_.........._11287.28__........._352.73_

RealLife-60%Rand-65%Read......_5.1278_..........__7639.49_.........__59.68___

Max Throughput-50%Read.........._7.5790_..........____7728_.........__241.53_

Random-8k-70%Read................._4.9714_.........._7852.49__........._61.35_

EXCEPTIONS:

##################################################################################

Backend io from the sanmelody performance monitor shows the SAS drive are easily capable of 2000 iops bursting up to 3000 ( probaby from controller cache)

Message was edited by:
dave01

Yeah looks pretty good. Initialy I coudlnt get more then 200 iops at any setting, took me a while to find there was a bad gbic on the FC switch.

Sure makes the traditional SAN's look bad, concidering that the SANmelody server is also doing AIM replication over IP, and will be doing this interstate eventually for DR purposes.

Just setting up that kind of replication using EMC/Hitachi etc would have cost us more thatn the datacore software alone.

I'll post up the results from the other environment we are building when its ready ( two HDS ams 500 with SanMelody ontop)

By the way there was no other load on the SAN during that test