Ask the Expert: VMware and VNX Integration Best Practices

Welcome everyone, this ATE session has began. Our experts are now ready to answer any question you post on this thread. Enjoy!

Perhaps there's a story you want to share with us, so that we can think in a certain direction and start asking questions? When I did ATEs a couple of years back, I started the ATE with a sort of kick-off to get it going. I like a sort of roadmap / planning so over the next few days or weeks I have some direction where tha ATE can go to.

Is there a specific topic you were hoping for? Or any developments you want to share? What about vVols and VSANs? Is there something you want to share with the storage dudes / dudettes out here?

Hello RRR thanks for your recommendations and to share your experience of the events with me.

Right now I am working a lot with Configuration guidelines for VMware multipathing on clarion/VNX Storage arrays and the best practices to provide the tools to configure the correct multipath policy on the VMWare host according to the Storage attached to it. And also I am working a lot with best practices to implement critical SQL/ORACLE servers over VNX or VMAX too and the best practices to applied on all the virtual environments with this kind of implementations.

Please if anyone has any questions or want to share information about this topics or other topics please be welcome.

Thanks & Regards

Alejandro

A couple of months ago we ran into an issue where RDMs provided to an SQL cluster were very slow. Everything looked good: failover mode 4 on the VNX side and round robin on the VMware side. For RDMs the recommendation however was to set the FO policy to 1 and use MRU (or fixed, I'm not entirely sure anymore). The performance of the disks went up and the customer was happy again. I was wondering if you've encountered more database server specific settings that could improve performance. for example: what if a small VNX 5200 is only used for SQL 2008 and 2012, is there some tweaking we can do on the VNX cache or disk allignment or number of paths from the host to the array. Is 4 paths on VNX always the best, then why is 2 the best practice when used on a VMAX? Do these settings have something to do with VMware. And what about queue depth, execution throttle and perhaps other settings on the HBAs of a host. Are there any VMware specific tweaks concerning this queue depth that we need to consider?

Hello RRR , for me from the begining is to not have RDM for problems with DRS , backups , and the rest of the know issues for RDM.

The best practices for me to implement on VNX is to have failover mode 4 ( ALUA ) with Fixed for the policy path selection by the side of VMware. On VMAX we are using as failover mode Active/Active form the side of the VMAX and Round Robin by the side of VMware.

To improve the performance of the VMs that runs DBs the best practices to implement are many for example:

1 - You can put for the disks of the vms when the DBs are allocated and all the disk on the vm Paravirtualized adapters for SCSI PVSCSI.

2 - Its very important that you format the vmware disks with  thick Provision Eager Zeroed to have a better performance by the side of the DBs.

3- Place binary SQL Server, log, and data in separate VMDK. In addition to better performance, separate them provides better flexibility to backup OS / SQL Server binary VMK can be backed up with snapshot-based backup and data recovery with vmware. The SQL Server data and log files can be guarded with backup solutions based traditional data.

4- If you have the chance to use powerpath its the best option to manage the multipath history.

5- Typical SQL Server Disk Access Patterns:

Operation                   Random/Sequential                       Rear/Write                             Size Range

OLTP - Log                     Sequential                                   Write                                     Up to 64K

OLTP - Data                    Random                                     Read/Write                             8k

Bulk Insert                       Sequential                                  Write                                      Any multiple of 8k up to 256k

Read Ahead (DSS,           Read                                          Read                                     any multiple of 8kb up to 512k     Index Scans)

Backup                            Sequential                                   Read                                      1MB

We have a lot of best practices to implement by the side of the VNX/VMAX , VMware and the DBs world too but these are a resume of the best for me.

If you need more specific information  please do not hesitate to contact me.

Thanks & Regards

Alejandro  

I'm a bit confused and I may just not be reading things right. You seem to suggest that PowerPath offers the best multipathing but you are recommending a fixed pathing policy on the VMware side of a VNX connection.

We have had some major problems with the fixed pathing because it seems to unbalance the load on the front end ports of the array that are used for ESX connectivity.

Hello Allen.

If you have the possibility to buy licenses of Powerpath its the best option recommended by EMC to manage multipathing on a VMware environment.

PowerPath can effectively manage all of the I/O multi-pathing complexity. It constantly adjusts the I/O path usage to changes in loads generated by the VMs. Management is vastly simplified. Just assign all devices (LUNs) to all available paths and PowerPath will optimize the overall I/O performance for the ESXi environment.

If needed, you can choose to provide additional QoS management for the most important applications by managing LUN and path prioritization.

If you don't have the possibility to use powerpath for license problems or another problems on your infrastructure Fixed its the best option for ESXI for VNX if you configure the trespassing with the correct paths by the side of ESXI you shouldn't have problems.

Do not hesitate to contact me if you have any questions or concerns.

Thanks and Regards.

Alejandro

This Ask the Expert event has concluded. But the conversation can continue as a regular Support Forum discussion.

Thanks to our experts for hosting and thanks to those who interacted ad asked questions. You can continue to ask your questions and conversing on this topic. 

Thanks again to all participants for your continued support of the Ask the Expert program.

Best regards,

Mark