VMware Cloud Community
christianZ
Champion
Champion

New !! Open unofficial storage performance thread

Hello everybody,

the old thread seems to be sooooo looooong - therefore I decided (after a discussion with our moderator oreeh - thanks Oliver -) to start a new thread here.

Oliver will make a few links between the old and the new one and then he will close the old thread.

Thanks for joining in.

Reg

Christian

574 Replies
abirhasan
Enthusiast
Enthusiast


IOpsMBpsLatency% CPU Utilization
Max Throughput-100%Read6774211.78.810.8
RealLife-60%Rand-65%Read391730.611.721.3
Max Throughput-50%Read6896215.58.710.6
Random-8k-70%Read349627.312.523.1

Latency  numbers cut in half basically. I read somewhere (but maybe not  in this  forum), to use as many workers in IOMETER as cpus & cores  in the VM,  but I'm guessing based in other numbers posted, that people  are only  using 1 worker

abirhasan   
0 Kudos
alexxdavid
Contributor
Contributor

Hi there

Needs some advice regarding my setup

Starting with the Servers :

2 x Dell R710 with quad port each reserved for iSCSI

48GB Ram Each

San:

Dell MD 3220i setup as :

2 x Controllers with 2GB Cache each

8 x 300GB 10 Krpm in Raid 10 in slot 1-8 owned by controller 0

8 x 146GB 15 Krpm in Raid 10 in slot 17-24 owned by controller 1

No Hot Spare

2 LUNS :

1 x 1.1TB ( the 8 x 300GB )

1 X 550GB ( the 8 x 146GB )

Now as you can see i have space in the san to put another 8 SAS Drive in slot 9-16

What would be the better solution from those :

1 : Buy 8 X 3000GB and create another lun with 1.1TB and owned it by another controller = 3 Luns total, one controller will own 2 luns

2 : Buy 4 X 300GB and 4 X 146GB and expand both the raid 10 already in place  = Same 2 Luns, 1 luns per each controller

3 : Buy 4 X 300GB and 4 X 146GB and expand both raid 10 but create 4 Luns, 2 luns per each raid 10, 2 luns per controller

Space is not an issue as total data for the business is under 1TB and right now, we still got massive space left, we have around 10 virtual machines.

Right now, both luns can achieve ~ 12000 - 13000 iops from iometer 100% read from 4 network card and 250Mbps from 50%read

Tell me your thoughts.

Regards

0 Kudos
dbrinkmann99
Enthusiast
Enthusiast

OpenPerformanceTest32 against a Nimble CS240, not bad for a bunch of 7.2K drives with some SSD for read cache Smiley Happy

SERVER TYPE:Windows 2008 R2 VM CPU TYPE / NUMBER: 5620 HOST TYPE: Cisco C200 M1 STORAGE TYPE / DISK NUMBER / RAID LEVEL: Nimble CS240 / Hybrid
Test name Latency Avg iops Avg MBps cpu load
Max Throughput-100%Read15.8937791180%
RealLife-60%Rand-65%Read4.44128981001%
Max Throughput-50%Read11.5050271570%
Random-8k-70%Read3.78153431190%

Old style VMTN communities table:

SERVER TYPE:Windows 2008 R2 VM
CPU TYPE / NUMBER: 5620
HOST TYPE: Cisco C200 M1
STORAGE TYPE / DISK NUMBER / RAID LEVEL: 
Nimble CS240 / Hybrid
|*TEST NAME*|*Avg Resp. Time ms*|*Avg IOs/sec*|*Avg MB/sec*|*% cpu load*|
|*Max Throughput-100%Read*|15.89|3779|118|0%|
|*RealLife-60%Rand-65%Read*|4.44|12898|100|1%|
|*Max Throughput-50%Read*|11.50|5027|157|0%|
|*Random-8k-70%Read*|3.78|15343|119|0%|
Dan Brinkmann
0 Kudos
zgz87
Contributor
Contributor

Hi everyone!

I would like to have your opinion about my results. I feel that I am doing something incorrectly since the MBps in the RealLife case and in the Random case are dramatically lower. Can this be possible? Also, is it normal to get the CPU utilization in all the tests equal to 0?

CPU type: Intel Xeon L5638 HC 2GHz (x2)

Host type: ESXi 5 / 96 Gb RAM / OpenSUSE 12.1 Kernel 3.1.0

Storage type: EMC2 VNX

Disk type: Pool of 30 x 15K550GB / RAID 5

LUN: 500 GB

Interface: FCoE SW driver (from ESXi vmware): ixgbe


              IOps          MBpsLatency% CPU Utilization
Max Throughput-100%Read3434107170
RealLife-60%Rand-65%Read51641150
Max Throughput-50%Read279087210
Random-8k-70%Read41631440

To end, I posted some questions about IOmeter in another thread that maybe someone could answer : http://communities.vmware.com/thread/397984

Thanks

0 Kudos
Ingens
Contributor
Contributor

Hi at all,

You can compare these reports between both storages below;

EMC VNXe 3100:

'Test Type,Test Description
0,IO-Test
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
TABLE oF RESULTS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++


SERVER TYPE: VM
CPU TYPE / NUMBER: VCPU / 1
HOST TYPE: CISCO UCS 200 M, 24GB RAM; 6 CPUs x Intel Xeon CPU E5645 2.40 Ghz
STORAGE TYPE / DISK NUMBER / RAID LEVEL: EMC VNXe 3100 / 6 Disks (7.200 RPM) x 868.961 GB / RAID 6


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


Max Throughput-100%Read......___21.40____......._2825.00__........._88.28___


RealLife-60%Rand-65%Read..___93.48____.........__562.42__...........__4.39___


Max Throughput-50%Read........___15.76____.......__3766.27__........._117.69___


Random-8k-70%Read...............___118.96____.......__440.12__.........__3.43___


EXCEPTIONS: CPU Util. 32% - 15% - 18% - 15%;

SUN SERIES 7110:


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

'Test Type,Test Description
0,IO-Test
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
TABLE oF RESULTS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++


SERVER TYPE: VM
CPU TYPE / NUMBER: VCPU / 1
HOST TYPE: HP DL320 G5p, 8GB RAM; 4 CPUs x Intel Xeon X3320 2.50 Ghz
STORAGE TYPE / DISK NUMBER / RAID LEVEL: SUN STORAGE 7110 / 16 Disks (10.000 RPM) x 137 GB / Double Parity


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


Max Throughput-100%Read......___21.71____......._2770.693__........._86.58___


RealLife-60%Rand-65%Read..___84.02____.........__699.9__...........__5.47___


Max Throughput-50%Read........___15.56____.......__3735.425__........._116.73___


Random-8k-70%Read...............___77.47____.......__741.82__.........__5.79___


EXCEPTIONS: CPU Util. 32% - 15% - 18% - 15%;


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

Please, could you provide feedback about the results achived?

Ingens,

Best regards

0 Kudos
oparcollet
Enthusiast
Enthusiast

SERVER TYPE: VM
CPU TYPE / NUMBER: VCPU / 2
HOST TYPE: DELL R610, 64GB RAM; 2 CPUs x Intel Xeon CPU E5540 2.37 Ghz
STORAGE TYPE / DISK NUMBER / RAID LEVEL:  ORACLE PILLAR AX600 r3 - AxiomOne 5.3 / 180 Disks (7.2K RPM) x 512 GB / RAID 5 / QOS : LOW (4  RAID GROUPs)

##################################################################################
TEST NAME--   Av. Resp. Time ms Av. IOs/sek    Av. MB/sek-
##################################################################################
Max Throughput-100%Read......___4,0021____......._14316,08__........_447,37___
RealLife-60%Rand-65%Read.....___1,632_____......._27211,02__......._213,67___
Max Throughput-50%Read........___2,2343____......._22773,40__........_710,83___
Random-8k-70%Read...............__1,6974_____......._24757,88__........_194,68___

SERVER TYPE: VM
CPU TYPE / NUMBER: VCPU / 2
HOST TYPE: DELL R610, 64GB RAM; 2 CPUs x Intel Xeon CPU E5540 2.37 Ghz
STORAGE TYPE / DISK NUMBER / RAID LEVEL:  ORACLE PILLAR AX600 r3 - AxiomOne 5.3 / 180 Disks (7.2K RPM) x 512 GB / RAID 5 / QOS : PREMIUM (30 RAID GROUPs)

##################################################################################
TEST NAME--   Av. Resp. Time ms Av. IOs/sek    Av. MB/sek-
##################################################################################
Max Throughput-100%Read......___3,8216____......._15547,60__........._484,32___
RealLife-60%Rand-65%Read....___1,8760____........_29324,05__........._229,83___
Max Throughput-50%Read.......___1,8841____........_28894,64__........._909,73___
Random-8k-70%Read..............___1,7808_____......_31189,50___......._243;60___

Update : HBA QLA 2462, qdepth=64, The SAN was running 400 others VMs (100 Servers + 300 View) at the same time. CPU usage 1 to 3%.

Enjoy!

0 Kudos
blvsupport
Contributor
Contributor

SERVER TYPE:Windows 2008 R2 64 Bits with 4 GB VM on a lun with 5 other VM
CPU TYPE / NUMBER: Intel Xeon x5650 / 4 HOST TYPE: IBM BladeCenter HS22 7870 STORAGE TYPE / DISK NUMBER / RAID LEVEL: IBM DS3512 with 2 Brocade 8Gbps / 48 hard disk Speed: 15000 RPM Data Rate 6Gbps/ 5
Test nameLatencyAvg iopsAvg MBpscpu load
Max Throughput-100%Read2.78214056680%
RealLife-60%Rand-65%Read10.403651280%
Max Throughput-50%Read4.46134294190%
Random-8k-70%Read10.283392261%
Max Throughput-100%Read2.76215096720%
RealLife-60%Rand-65%Read10.563655281%
Max Throughput-50%Read4.39135464230%
Random-8k-70%Read10.383419261%

Is it normal like result what do you think ?

Ive done the test 2 times with about the same result

Thanks !

0 Kudos
zgz87
Contributor
Contributor

Pretty good numbers. In my opinion they are in a range which you can consider as normal but I would need more info. How are you accessing the storage (iSCSI, NFS...)? How is the link to the storage (1GbE, 10GbE...)? Does your server have a specific storage interface or are you using a software driver?

0 Kudos
blvsupport
Contributor
Contributor

Cool Thanks

The blade chassis is link to the storage IBM DS3512 by 2 Fiber Channel Brocade 8 Gbps

Each FC card have 2 links to the storage

The IBM DS3512 is link to my 3 expansion by SAS cable

Hope i give enough details

Thanks !

0 Kudos
MKguy
Virtuoso
Virtuoso

Your numbers don't seem bad, but I would expect a bit better performance for the random access heavy tests with 48 15k RPM disks.

Is this ESXi5 U1? Btw. one vCPU for the test VM should be sufficient and is it using pvsci?

Is there significant other normal IO going on on the array besides your benchmark and could you put the test VM on it's own LUN?

Also, are you using the roundrobin path selection policy and tried setting the IOPS switchover parameter to 1?

(See esxcli http://pubs.vmware.com/vsphere-50/index.jsp?topic=%2Fcom.vmware.vcli.examples.doc_50%2Fcli_advanced_...

like "esxcli storage nmp psp roundrobin deviceconfig set --type "iops" --iops 1 --device [device ID goes here]")

-- http://alpacapowered.wordpress.com
0 Kudos
mry
Contributor
Contributor

Jeg arbejder ikke for styrelsen mere.

Kontakt drift@eogs.dk eller +45 3330 7587

I do not work for the agency anymore.

Please contact drift@eogs.dk or 0045 33307587

0 Kudos
joshrobinson
Contributor
Contributor

SERVER TYPE:Dell PowerEdge R620 CPU TYPE / NUMBER: Intel Xeon E5-2665 @ 2.4GHz, 1 HOST TYPE: Windows Server 2003 R2 32-bit STORAGE TYPE / DISK NUMBER / RAID LEVEL: Equallogic (2)PS6000 (1)PS5000, 48 Disks, Raid 10

Test name

LatencyAvg iopsAvg MBpscpu load
Max Throughput-100%Read8.16723522618%
RealLife-60%Rand-65%Read23.2618391445%
Max Throughput-50%Read28.3020656414%
Random-8k-70%Read23.3218481444%

The sans are in production currently so I dont know if that is throwing my numbers off.  Also, we just put in some new Force10 S50 switches but it appears that our Latency is high?  What do you guys think?

Josh

0 Kudos
davidarnold
Contributor
Contributor

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
TABLE OF RESULTS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
SERVER TYPE: Ubuntu 12.04 VM on ESXi 4.1
CPU TYPE / NUMBER: VCPU / 1
HOST TYPE: Dell PET410, 6GB RAM; 4x XEON E5503, 2.00 GHz
STORAGE TYPE / DISK NUMBER / RAID LEVEL: Dell PE2900 x 1 / 2+1 SAS10k / R5
SAN TYPE  / HBAs : iSCSI
\##################################################################################
TEST NAME--
Av. Resp. Time ms--Av. IOs/sek---Av. MB/sek----
\##################################################################################
Max Throughput-100%Read........__1.28______..........___777___.........___24____
RealLife-60%Rand-65%Read......___4.52_____..........___221___.........____1.7____
Max Throughput-50%Read..........____1.97____..........___506___.........___15____
Random-8k-70%Read.................____4.92____..........___203___.........____1.59____
Can I just query the best way to run this test? I have 1 worker and run the first test, checked result, second test, checked result etc. Should I be running 4 workers at the same time or another setup?
Thanks
David
0 Kudos
darkbgr123
Contributor
Contributor

I haven't seen many other results from the VMware VSA virtual appliance, should I be happy?

I have 2 hosts, and 8x 300GB SAS drives on each host. on the p410i i have 75%write and 25% read setting, and RAID6 setup, for a total of 1.6TB usable VSA storage.

Are my numbers looking decent enough you think?

Thanks for any input!

8GB test results:

SERVER TYPE: Win2008 x64 VM running on Proliant DL380 host (part of VSA cluster) CPU TYPE / NUMBER: 2 vCPUs E5605 HOST TYPE: Proliant DL380 G7 STORAGE TYPE / DISK NUMBER / RAID LEVEL: RAID6 on the P410i Array , 8x SAS 10k RPM drives, and software RAID0 from the VSA Appliance from VMware (Software NAS)
Test name Latency Avg iops Avg MBps cpu load
Max Throughput-100%Read0.00204116372%
RealLife-60%Rand-65%Read1.1239230%
Max Throughput-50%Read97.0159171840%
Random-8k-70%Read0.8333920%

Old style VMTN communities table:

SERVER TYPE: Win2008 x64 VM running on Proliant DL380 host (part of VSA cluster)
CPU TYPE / NUMBER: 2 vCPUs E5605
HOST TYPE: Proliant DL380 G7
STORAGE TYPE / DISK NUMBER / RAID LEVEL: RAID6 on the P410i Array , 
8x SAS 10k RPM drives, and software RAID0 from the VSA Appliance from VMware (Software SAN)

|*TEST NAME*|*Avg Resp. Time ms*|*Avg IOs/sec*|*Avg MB/sec*|*% cpu load*|
|*Max Throughput-100%Read*|0.00|20411|637|2%|
|*RealLife-60%Rand-65%Read*|1.12|392|3|0%|
|*Max Throughput-50%Read*|97.01|5917|184|0%|
|*Random-8k-70%Read*|0.83|339|2|0%|

32GB Test results

SERVER TYPE: Win2008 x64 VM CPU TYPE / NUMBER: 2 vCPUs E5605 HOST TYPE: Proliant DL380 G7 part of VSA cluster STORAGE TYPE / DISK NUMBER / RAID LEVEL: RAID6 on the P410i Array , 8x SAS 10k RPM drives, and software RAID0 from the VSA Appliance
Test name Latency Avg iops Avg MBps cpu load
Max Throughput-100%Read0.00214486702%
RealLife-60%Rand-65%Read1.1138730%
Max Throughput-50%Read99.0060401880%
Random-8k-70%Read0.8333720%

Old style

SERVER TYPE: Win2008 x64 VM
CPU TYPE / NUMBER: 2 vCPUs E5605
HOST TYPE: Proliant DL380 host (part of VSA cluster)
STORAGE TYPE / DISK NUMBER / RAID LEVEL: RAID6 on the P410i Array ,
8x SAS 10k RPM drives, and software RAID0 from the VSA Appliance |*TEST NAME*|*Avg Resp. Time ms*|*Avg IOs/sec*|*Avg MB/sec*|*% cpu load*| |*Max Throughput-100%Read*|0.00|21448|670|2%| |*RealLife-60%Rand-65%Read*|1.11|387|3|0%| |*Max Throughput-50%Read*|99.00|6040|188|0%| |*Random-8k-70%Read*|0.83|337|2|0%|
0 Kudos
jones_chris
Contributor
Contributor

Seeing as its taken me way to long to finally get a decent round robin performance from 2 gig links, thought I'd share them for posterity.
Turns out that FreeNAS/Nas4Free just hasn't got a good enough iscsi target, OpenFiler does.

SERVER TYPE: HP N36L Microserver CPU TYPE / NUMBER:  AMD Athlon II Neo N36L, 4GB RAM, running OpenFiler 2.99 iSCSI target with BlockIO HOST TYPE: HP ML110 G7, 8GB RAM, Windows Server 2008 R2. 2 Gig links running Round Robin on separate VLANs
  to N36L with 2 Gig links through 3COM switch. STORAGE TYPE / DISK NUMBER / RAID LEVEL: 1 x 60GB OCZ Vertex SSD.
Test nameLatencyAvg iopsAvg MBpscpu load
Max Throughput-100%Read0.0062691950%
RealLife-60%Rand-65%Read19.906936540%
Max Throughput-50%Read76.4946661450%
Random-8k-70%Read13.385444420%
0 Kudos
davidarnold
Contributor
Contributor

Hi Chris,

When you are using Openfile are you not concerned about: http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=102659... ?

David

0 Kudos
jones_chris
Contributor
Contributor

davidarnold wrote:

Hi Chris,

When you are using Openfile are you not concerned about: http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=102659... ?

David

It was more of a proof of concept as I was struggling to get the performance I knew should be possible. In reality, we have 10G from a Thecus N8800Pro to our physical (non VM) server which is the only one using iSCSI. The ESXi machines only use local storage for (running) VM's.

0 Kudos
Darking
Enthusiast
Enthusiast

Not sure if anyone has allready posted Equallogic PS6100XS hybrid array stats, but here goes.

SERVER TYPE: M805 CPU TYPE / NUMBER: 2 Vcpu Opteron 2376 HOST TYPE: Windows 2008 R2 on VMware Via Citrix-PVS STORAGE TYPE / DISK NUMBER / RAID LEVEL: Equalogic PS6100XS
Test name Latency Avg iops Avg MBps cpu load
Max Throughput-100%Read0.00725122629%
RealLife-60%Rand-65%Read18.226362494%
Max Throughput-50%Read129.56791224731%
Random-8k-70%Read16.306630514%
0 Kudos
benoit10
Contributor
Contributor

darking, you have 17 x 600 + 4 x 400 SSD , right ?

0 Kudos
Darking
Enthusiast
Enthusiast

7 SSDs 400 gig and 17 600 gig 10K Sas drives running raid 6 Accelerated

Around 13TB usable capacity

0 Kudos