VMware Communities: Open unofficial storage performance thread

Re: Open inofficial storage performance thread Sep 6, 2007 7:15 PM

chucks0 88 posts since
Feb 2, 2005

I just finished benchmarking our EQ PS300 and got some confusing results. It appears that using the MS iSCSI initiator from within a VM is considerable faster than using the ESX Software initiator. How can I improve the software initiator? Also, it appears that the results from IOMeter for the third test with the ESX Software initiator are not correct.

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

SERVER TYPE: Win2k3 VM (MS iSCSI initiator in VM)
CPU TYPE / NUMBER: VCPU / 1
HOST TYPE: Dell PE2900, 24GB RAM; 2x XEON 5160, 3.00 GHz, DC
STORAGE TYPE / DISK NUMBER / RAID LEVEL: EQ PS300E x 1 / 12 Disks (SATA7.2k) / R50

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

Max Throughput-100%Read........___19.72__..........___3014___.........___94.19___

RealLife-60%Rand-65%Read......___23.57__..........___1800___.........___14.06__

Max Throughput-50%Read..........___22.43__..........___2436___.........___76.14__

Random-8k-70%Read.................___26.10__..........___1645___.........___12.85___

EXCEPTIONS: CPU Util.-85,65,88,61;

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

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

SERVER TYPE: Win2k3 VM (ESX iSCSI initiator)
CPU TYPE / NUMBER: VCPU / 1
HOST TYPE: Dell PE2900, 24GB RAM; 2x XEON 5160, 3.00 GHz, DC
STORAGE TYPE / DISK NUMBER / RAID LEVEL: EQ PS300E x 1 / 12 Disks (SATA7.2k) / R50

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

Max Throughput-100%Read........__27.56___..........___2154___.........___67.31___

RealLife-60%Rand-65%Read......___51.16__..........___1109___.........___8.66___

??Max Throughput-50%Read..........__10.65___..........___5136___.........___160.52___

Random-8k-70%Read.................___58.16__..........___982____.........___7.67____

EXCEPTIONS: CPU Util.-27,23,64,22;

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

Re: Open inofficial storage performance thread Sep 7, 2007 4:18 AM

in response to: chucks0

pops106 218 posts since
Sep 7, 2004

Just some more results;

SERVER TYPE: VM
CPU TYPE / NUMBER: VCPU / 1
HOST TYPE: HP DL585, 16GB RAM, 4 x 2.4ghz Single core
STORAGE TYPE / DISK NUMBER / RAID LEVEL: DotHill SATA / 12 / Raid 5 / 250gb Disks / 10gb VMDK on 250GB vmfs LUN

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

Max Throughput-100%Read........13.6071..........4317.64........134.93

RealLife-60%Rand-65%Read......108.0142..........495.41.........3.87

Max Throughput-50%Read..........12.5936..........4503.01.........140.72

Random-8k-70%Read.................134.4153..........409.6.........4.00

CPU = 30% - 28% - 59% - 25%

Re: Open inofficial storage performance thread Sep 7, 2007 4:57 AM

in response to: pops106

pops106 218 posts since
Sep 7, 2004

SERVER TYPE: VM
CPU TYPE / NUMBER: VCPU / 1
HOST TYPE: HP DL585, 16GB RAM, 4 x 2.4ghz Single core
STORAGE TYPE / DISK NUMBER / RAID LEVEL: DotHill FC / 12 / Raid 5 / 144gb 10krpm Disks / 10gb VMDK on 250GB vmfs LUN

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

Max Throughput-100%Read........13.4877.........4350.52........135.95

RealLife-60%Rand-65%Read......50.6972..........1020.62.........7.97

Max Throughput-50%Read..........11.1034..........5192.04.........162.25

Random-8k-70%Read.................56.7146..........925.31.........7.23

CPU = 36% - 34% - 58.35% - 31.96%

Re: Open inofficial storage performance thread Sep 7, 2007 7:54 AM

in response to: pops106

pops106 218 posts since
Sep 7, 2004

SERVER TYPE: VM
CPU TYPE / NUMBER: VCPU / 1
HOST TYPE: HP DL585, 16GB RAM, 4 x 2.4ghz Single core
STORAGE TYPE / DISK NUMBER / RAID LEVEL: NetAPP FAS3020c / 14 / Raid DP / 250gb SATA Disks / 10gb VMDK on 500GB vmfs LUN

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

Max Throughput-100%Read........10.1373.........5682.14........177.57

RealLife-60%Rand-65%Read......41.2474..........1341.37.........10.48

Max Throughput-50%Read..........7.4528..........7354.47.........229.83

Random-8k-70%Read.................61.9947..........903.06.........7.06

CPU = 29% - 25% - 70%% - 23%

Re: Open inofficial storage performance thread Sep 7, 2007 8:57 AM

in response to: chucks0

Atamido 34 posts since
May 31, 2007

@chucks0:

It appears that using the MS
iSCSI initiator from within a VM is considerable
faster than using the ESX Software initiator.

I know that was a pretty normal situation with ESX 3.0.1 and before, but I don't know about 3.0.2. What version are you running in those tests?

Re: Open inofficial storage performance thread Sep 7, 2007 9:39 PM

in response to: Atamido

chucks0 88 posts since
Feb 2, 2005

I have tested with both 3.01 and 3.02 and get the same results.

Message was edited by:
chucks0

Re: Open inofficial storage performance thread Sep 8, 2007 12:54 AM

christianZ 2,212 posts since
Apr 21, 2006

Wow !! So many new results!

I would to thank to

multirotor
chucks0 (Yes it was the first midrange EMC test!)
pops106

for your tests.

@ chucks0
Yes that's true, the Esx iscsi initiator is very poor; even with iscsi hba you can't reach so high number like by ms iscsi initiator (it is a ESX problem not EQL). When you will check my results (e.g) and results from guys with FC, you will see that ISCSI in ESX doesn't run so effective like FC (differences between physical and virtual).

@ pops106
By your Netapp are you configured one 14-disks aggregate or 2x 7 ?

Re: Open inofficial storage performance thread Sep 12, 2007 10:41 AM

sbelisle 2 posts since
Sep 12, 2007

4-5 VM's assigned to physical hardware, but sitting idle during test. Disk presented as single R5 LUN (200GB) to the ESX server. VM instance IO ran on had 512MB assigned memory. VMDK 20gb C:, 180GB D:

SERVER TYPE: VM or PHYS. - VM ESX 3.0.1
CPU TYPE / NUMBER: VCPU / 1
HOST TYPE: DL380G4 2x3.2GHZ DC XEON, 24GB RAM
STORAGE TYPE / DISK NUMBER / RAID LEVEL: CX700 / 4+1 Disks R5 / 500gb ATA 7200RPM

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

Max Throughput-100%Read........___10.79____..........____4920__.........____153.7___

RealLife-60%Rand-65%Read......___79.43____.........._____692__........._____5.4___

Max Throughput-50%Read..........___7.78____..........____5361__.........___167__

Random-8k-70%Read.................__60.51_____..........__467____........._____3.6___

EXCEPTIONS: CPU Util.82,30,90,27%;

Message was edited by:
sbelisle

Re: Open inofficial storage performance thread Sep 12, 2007 10:53 AM

sbelisle 2 posts since
Sep 12, 2007

Same test as above but to a 4x50=200GB meta

SERVER TYPE: VM or PHYS. - VM ESX 3.0.1
CPU TYPE / NUMBER: VCPU / 1
HOST TYPE: DL380G4 2x3.2GH DC XEON, 24GB RAM
STORAGE TYPE / DISK NUMBER / RAID LEVEL: CX700 / 4x4+1 Disks R5 / 500g ATA

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

Max Throughput-100%Read........___10.98____..........____5043__.........____157.5___

RealLife-60%Rand-65%Read......___40.59____.........._____1203__........._____9.4___

Max Throughput-50%Read..........___10.63____..........____5131__.........___160.3__

Random-8k-70%Read.................__44.70_____..........__1128____........._____8.8___

EXCEPTIONS: CPU Util. 73,43,79,40

Re: Open inofficial storage performance thread Sep 13, 2007 10:59 PM

in response to: sbelisle

christianZ 2,212 posts since
Apr 21, 2006

Thanks for joining in.

Re: Open inofficial storage performance thread Sep 17, 2007 4:25 AM

D.duke 2 posts since
Jul 6, 2006

Test was the only thing running on the system (not yet in production)
testfilesize set to 20000000 (10 GB file)
Tests were run on a VRaid0,1 and 5 Vdisk (size Vdisk was 100 GB in all cases)

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
TABLE OF RESULTS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

SERVER TYPE: VM ON ESX 3.0.2
CPU TYPE / NUMBER: VCPU / 1
HOST TYPE: HP ProLiant BL460c G1, 16GB RAM; 2x Intel XEON 5345, 2,33 GHz
STORAGE TYPE / DISK NUMBER / RAID LEVEL: HP EVA 6100 / 40 Fiber Channel 15k / Vraid1
SAN TYPE / HBAs : Fiber, Emulex LPe 1105-HP HBA

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

Max Throughput-100%Read........__0,61____..........___6210___.........___194____

RealLife-60%Rand-65%Read......___6,1____..........___3667___.........____29____

Max Throughput-50%Read..........____6,2___..........___4336___.........___136____

Random-8k-70%Read.................____5,7___..........___3579___.........____28____

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

SERVER TYPE: VM ON ESX 3.0.2
CPU TYPE / NUMBER: VCPU / 1
HOST TYPE: HP ProLiant BL460c G1, 16GB RAM; 2x Intel XEON 5345, 2,33 GHz
STORAGE TYPE / DISK NUMBER / RAID LEVEL: HP EVA 6100 / 40 Fiber Channel 15k / Vraid5
SAN TYPE / HBAs : Fiber, Emulex LPe 1105-HP HBA

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

Max Throughput-100%Read........__0,56_____..........___8605___.........___269____

RealLife-60%Rand-65%Read......___9,3____..........___5113___.........____40____

Max Throughput-50%Read..........____38,3__..........___1374___.........___43____

Random-8k-70%Read.................____9,1___..........___5242___.........____41____

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

SERVER TYPE: VM ON ESX 3.0.2
CPU TYPE / NUMBER: VCPU / 1
HOST TYPE: HP ProLiant BL460c G1, 16GB RAM; 2x Intel XEON 5345, 2,33 GHz
STORAGE TYPE / DISK NUMBER / RAID LEVEL: HP EVA 6100 / 40 Fiber Channel 15k / Vraid0
SAN TYPE / HBAs : Fiber, Emulex LPe 1105-HP HBA

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

Max Throughput-100%Read........__0,54____..........__10068___.........___314____

RealLife-60%Rand-65%Read......___7,3____..........___6928___.........____54____

Max Throughput-50%Read..........____6,8___..........___5430___.........___170____

Random-8k-70%Read.................____6,8___..........___7476___.........____58____

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

Re: Open inofficial storage performance thread Sep 18, 2007 11:02 AM

in response to: D.duke

christianZ 2,212 posts since
Apr 21, 2006

Wow. That's really not bad. Thanks for testing.

Re: Open inofficial storage performance thread Sep 24, 2007 8:22 PM

in response to: chucks0

kenrobertson 9 posts since
Feb 1, 2007

Anyone done any tests with IET? Curious how IET's performance compares to SanMelody. I am currently looking at both... IET is attractive for being free, but SanMelody would likely have more support.

Re: Open inofficial storage performance thread Sep 24, 2007 8:44 PM

in response to: kenrobertson

thorwitt 34 posts since
Jan 24, 2007

Hallo,

ich bin vom 22.09.2007 bis zum 07.10.2007 nicht im Hause erreichbar.
Ihre Mail wird an Herrn Sachs weitergeleitet und bearbeitet.

Mit freundlichen Grüßen
TargoSoft IT-Systemhaus GmbH

Thorsten Witt
Technik
******************************************************************
***

TargoSoft IT-Systemhaus GmbH
Poseidonhaus
Amsinckstrasse 65
D-20097 Hamburg
Telefon 040 / 23 51 22-23
Telefax 040 / 23 51 22-40
http://www.targosoft.de
Geschäftsführer: Wolfgang Wündsch
Handelsregisternummer: HRB 33 548 Hamburg
Gerichtsstand: Hamburg

***
******************************************************************
Datensicherung . Archivierung . Hochverfügbarkeit . Konsolidierung
Virtualisierung ...im heterogenen SAN

Vereinbaren Sie einen Termin in unserem mehrfach ausgezeichneten
OPEN-SAN Solution Center(r) unter : http://opensan.TargoSoft.de

communities-emailer 25.09.2007 05:22 >>>

Thorsten Witt,

A new message was posted in the thread "Open inofficial storage performance thread":

http://communities.vmware.com/message/756867

Author : kenrobertson
Profile : http://communities.vmware.com/people/kenrobertson

Message:

Re: Open inofficial storage performance thread Sep 27, 2007 10:50 AM

cmanucy 177 posts since
Jun 29, 2006

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

SERVER TYPE: Win2K3 VM on ESX 3.0.1
CPU TYPE / NUMBER: VCPU / 1
HOST TYPE: HP DL385 G1, 16GB RAM, 2x 2.6GHz AMD Opteron Dual Core
STORAGE TYPE / DISK NUMBER / RAID LEVEL: HP MSA 1500 / MSA30 with 5 x 147G 15K SCSI / RAID5

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

Max Throughput-100%Read........____12____..........___4772___.........___149____
RealLife-60%Rand-65%Read.......____59____..........____815___.........____6_____
Max Throughput-50%Read.........____22____..........____2625__.........____82____
Random-8k-70%Read..............____40____..........___1111___.........____9_____

EXCEPTIONS: CPU Util.-49-38-35-45;

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

Note: about 5VM's on this same LUN, tests run during fairly low use time.

Actions

Incoming Links