VMware Cloud Community
AlltitudeDVince
Contributor
Contributor

Performances: IBM x3650 QLE4062c DS3300 C3560

Hello,

My company just bought:

2x IBM x3650 loaded with 2x HBA Qlogic 4062C (dual port), 2x 73GB 10000 on IBM 8K/8i (raid 1)

1x DS3300 loaded with 12x 500GB SATA 7200

2x Cisco C3560.

I have no major issue regarding the connections but I am very surprised about the performances:

There is no performances differences between RAID1 (2 HDD), RAID10 (4 HDD) and RAID10 (6HDD)

I also have tested 2x RAID5 (5HDD), each on a dedicated controller/HBA/PATH, there is no differences.

In term of performances, I just can reach 30 MBytes/sec wrinting on HDD (iSCSI or local RAID disks) ?!?!?!?!?

Is that normal ?

FYI:

The HBA / DS3300 are connected on physicaly dedicated switches (c3560) which are NOT connected to anything else.

Firmware are all up to date (according to the release available on IBM.com ).

Any idea how I can increase performances ?

Thanks to all in advance

0 Kudos
7 Replies
olliecampbell
Contributor
Contributor

Hi,

We have exactly the same issue, slow write performance to a RAID 5 array running ESX 3i. We're starting to run out of things to try as well.

Sorry I don't have more of an answer.

Ollie

0 Kudos
BenConrad
Expert
Expert

Are you guys using IOMeter for this testing, most people consider this the industry standard I/O testing tool. Tests like Robocopy file copies are not going to cut it in terms of stress testing the SAN.

Regarding the SANs, do you have write-back caching enabled on each array? Write caching on the arrays will improve performance dramatically.

Ben

0 Kudos
sscheller
Contributor
Contributor

Hi,

same problem with esx4 software iscsi, ds3300 with sas hdds.

Poor write performance.

0 Kudos
olliecampbell
Contributor
Contributor

Unfortunately our's was so bad we reverted to physical hardware Smiley Sad

0 Kudos
joshuatownsend
Enthusiast
Enthusiast

I was experiencing performance issues with my DS3300 and managed to resolve them. I wrote up the solution on my blog here: http://vmtoday.com/2009/06/ibm-ds3300-iscsi-write-performance-solved/ - not sure how applicable my fix is to your situation, but probably worth a look.

If you found this or other information useful, please consider awarding points for "Correct" or "Helpful". Please visit http://vmtoday.com for News, Views and Virtualization How-To's Follow me on Twitter - @joshuatownsend
0 Kudos
Rumple
Virtuoso
Virtuoso

In term of performances, I just can reach 30 MBytes/sec wrinting on HDD (iSCSI or local RAID disks) ?!?!?!?!?

If you are seeing the same performance to local disk you might want to ensure you are testing in a VM, not using service console writes. The service console is limited for disk and network writes so you will not get good performance.

Network performance and the storage performance won't effect the local disk. I am pretty sure the 3650's do have the BBWC on them so as long as the write back is enabled you should be getting great performance to the local disk.

Couple other things:

Have you configured jumbo frames on the network cards, switches (I believe 3560's areswitch leve, not port level changes), are those switches stacked (ie, using mode on etherchannel) or are you using LACP (which I belive is mode auto or active)

Have you used the virtual center client to create the vmfs volumes on the ds3300 vs the installer? This will align the partitions correctly.

Curiosity question - with the amount of money spent, why would you go with sata drives vs SAS. Obviously you get more space, but you are going to hit disk bottlenecks long before you get to the end of the performance on the rest of the environment.

0 Kudos
sscheller
Contributor
Contributor

Hi,

we found out from oure tests, that not write performance is the iusse, its the read performance.

Write performance is realy good.

Up tp 80 MB/s in vm guest.

We although have to think about performance local disks, if you use a laptop or desktop pc for iometer tests.

But the bottleneck is read performance.

Don´t know why.

greatings

0 Kudos