Performance Issues with VMware Environment

We have not resolved this issue. We have done every kind of test here is what we are done till now

1. Applied EMC Pack on top of 5875.249.188
2. Assigned Disk from SSD, FC, SATA and have the same kind of throughput when we copy data from one disk to other . But we see excellent throughput when we copy data from LAN to local disk on VMAX which is around 150 to 250 mb in compared to 20 to 30 mb locally.
3. The same hosts does well on clariion
4. We have assigned FAST policy to see if anything different
5. We have 8 FA ports assigned to ESX hosts
   

We are still waiting for EMC to check what is going on with the frame. We are literally seeing 200 to 300 ms latency on the vmax Fiber chanel disks.

Has any one see this kind of issues on Vmware. The UCS environment Works perfectly without changes in the Environment. As soon as the Storage VMare vmotioned to VMAX datastore we see the issues .

We know this not issue with the UCS infra as the same configuration works fine with Clariion.

VNX and VMAX have different software architectures.  Did you try changing the parameter that sets how many IOs are sent down a channel before switching?  Did you try more paths? 

What is the workload?  It is possible that the type of workload you are running is just faster on the VNX than the VMAX.

Hi Sasi,

Thanks for your email. We have Set these settings on the initiator groups.

Thanks and Regards,

Govind.N.R

Storage Engineer, Information Technology

WellCare Health Plans, Inc.

8735 Henderson Rd, Ren 1, Tampa, FL 33634 | Tel-WOrk: 813-206-5731 x5731 | BB: 8130351-9762 |email: rgovind1@wellcare.com

We are on round robin and our IO switch value is 1. Which means every 1 IO the ESX host will switch the path. This is recommended by EMC after we have the issue when the default value is 1000.

Thanks and Regards,

Govind.N.R

Storage Engineer, Information Technology

WellCare Health Plans, Inc.

8735 Henderson Rd, Ren 1, Tampa, FL 33634 | Tel-WOrk: 813-206-5731 x5731 | BB: 8130351-9762 |email: rgovind1@wellcare.com

We have 8 paths 4 per fabric to each host. Our current setting on the Host for the IO is round robin and it is set to 1 IO before it switches.

Thanks and Regards,

Govind.N.R

Storage Engineer, Information Technology

WellCare Health Plans, Inc.

8735 Henderson Rd, Ren 1, Tampa, FL 33634 | Tel-WOrk: 813-206-5731 x5731 | BB: 8130351-9762 |email: rgovind1@wellcare.com

What is the workload?  What is the performance expectations for the workload?  What do you see on VNX and on VMAX?  How many FAs on the VMAX?

Hi,

Hope all required port flags are set for VMware environment,

i) Common Serial Number (C)

ii) VCM State (VCM) --- (ACLX for V-MAX)

iii) SCSI 3 (SC3)

iv) SPC 2

v) Unique World Wide Name (UWWN)

vi) Auto-negotiation (EAN)

vii) Point to Point (P)

Thanks,

Sasi

Hi Govind,

Did you give a try using the path policy to Round Robin if it is in Fixed or MRU.

Thanks,

Sasi

I didn't see an answer to the question about the workload profile and performance expectation.

What is that Quincy you are looking for.

We have Good Sized SQL host with 8 to 10 TB VMS with bulk data read and writes withing ESX Farm.

Our work load is like 80% read and 20% write.

Thanks

Govind

Ok, just wondering if the workload was OLTP like, or DW like, or maybe something single threaded like file copy.  From the "bulk read and write" comment, it sounds like it could be a file copy type workload.  This is one where adding concurrency may help if possible.

Here is what our SQL hosts are doing

We have an APP that reads the data from one database and then make three times write on other 3 hosts with additional data that reads from files.

Our SQL Vms are located on different Meta devices on the VMax.

Thanks

govind

And what is the performance expectation?  We should see about 100-150MB/sec for a file copy on VMAX.