HPE Storage Users Group

Hi All,

I am newbie for 3Par,
I have 3Par 7400 4 controller , Hp Blade c7000 and Bl460c 256GB Ram Hosts.
72 900GB 10000 Rpm Sas Disks.

3par side :
CPG Raid5 (5 data + 1 parity)
2TB VV

Vmware side :
I created vm disk eagerd zeroed, 64K regular formatted (not quick format)
Path Selection is Round Robin

Questions:
1- What should i see I/O value of my 3Par. Are there any table for 3Par I/O values or how can i calculate with my item list disk, controller...etc?
2- I am trying look my I/O values via Sqlio program but i am not sure values are correct?

My problem is poor disk performance on my virtualization platform of Sql Server.

Thanks

Hi,

with that config you should see about 17.7K IOPS in random reads and 11k IOPS in an 80/20 read/write mix. These are front end numbers (as measured on host ports or aggregate Windows perfmon).

what is your sqlio command and output?

Thanks,

Maurice

[I work for HP. my answers here represent my own opinions based on my experience]

Yes he will get this iop on a physical host. But on virtual host he get approximatly half of this iops.
We make test our 7200 with 8*920GB ssd. At physical host we get 130K iops (70 read/30 wrtite) from 1 LUN. But when we install esx on that host, create virtual machine and make fc pass throught connection we get only 40-50K iops.
Maybe we curves hands, and there is something wrong done in the settings ESX.
Because of this, we were forced to resign from the virtualization server ORACLE DBMS.

Interesting. is this using a thin vv presented as a raw RDM (Physical) disk to the VM or RDM (Virtual)?

Both variants.

Did you configure custom claim rules, setting IOPS=1?
It can have a significant impact on some workloads...

Also if you are using the virtual connect module it is very sensitice to driver-firmware combinations.
I'd make sure to download latest async drivers and VC firmware to rule out any issues there.

Also guest OS settings does have an effect, using pvscsi over LSI SAS virtal device can have a 30% performance increase.
When you say FC pass through, do you mean raw device mapping?
If it is for SCSI sharing, you need physical compatability mode, and pvscsi might not be supported.

Using VC fc modules. all with latest firmware. Mezzanine with latest firmware. Yes raw device mapping.

What version of vSphere are you running? Is this a Microsoft SQL cluster?

I have tested a 7450-4 with 32x 2TB SSD.
We did not get more as 60k IOPS out of one ESX Server.
(2x CPU, 2x 8GB FC or 2x 20GB FcOE)
Using different Hardware and Vendors for the Server.

We are using also RAW unformated Disks with multiple IO Meter's in the VMs.
Also testing all ESX Tweeking with Buffers, Drivers and all these tings.

At the End we stopped at 8 ESX Server and got arround 450k IOPS at 0.7ms.
So the Storage or backend was not the Limit.

I don't see a limit on the vm side. on my tests (Using FIO), I got 450K iops @8k random read
from a single VM/single blade. From the disk side, thin vv passed through as physical RDM disk using ESX 5.5 u2

In order to drive the IO up, I used 8 jobs/files. numjobs (threads)=16, IOdepth = 4, so this is 512 reader threads on 8 files. Latency also in the 0.7 range

do a sanity check on the SQLIO parameters, it could be you are undriving the tool, and it may seem as a limit, but it's not. Increase files/thread/QueueDepth until you get good IOPS at a reasonable latency.

When using a SINGLE file, you will hit a wall. Also if you use small files, your results will be skewed upward, because you are getting 100% cache hit on your reads. in my case I had to bump the file size up, to 128GB for a total test spread of 1TB. (my cache is 32GB per node, 128GB Tot) so I look for a <10% cache to data ratio to be sure I am hitting the disks most of the time. check statcmp during your sqlio run to verify your iops are not just comming from cache.