EVA P6350 - Hyper-V - Fiber -ISCI

[Drexel480]

So my company is in the process of ordering new hardware. They want a EVA P6350. Now the intent is to create a Virtual Lab environment using Hyper-V from Windows 2012 so that we can create virtual Fiber connections on the Guest Hosts VM's and access the SAN.

We plan on creating multiple failover Clustered SQL Guests hosts. Each SQL cluster has 5 shared Disks that need to be built on the SAN. In total I thinking this will add up to around 100 logical Disks to accomodate all of the SQL clusters. Each of the logical disk only need to be about 10 Gig each in size. The P6350 is going to have 25 - 900Gig drives in it for about 27 TB of total space.

My question is if we are doing all fiber connections can we use Virtual Disks on the SAN is are they only accessible if you use ISCI Target and Hosts.

So a little more information:

If I take my 27 TB of space and I want to chop that up into 4 Arrays say: 10TB, 10TB, 6TB and 1TB

Can I use Virtual Disks to chop up the 1TB space into these 100+ drives I need for my SQL Guest Hosts and mount the virtual disk as local logical disk over Fiber or can this only be done over ISCI?

Second part of my question is if I cannot access Virtual disk over Fiber is there a pretty easy way to break up just 1 TB of the SAN into 100+  10GB logical disks so that I can access them over fiber connections?

I will apologize up front because SAN's are my weak point and so i hope I have explained my question correctly.

Thanks,

John

[Johan Guldmyr]

From http://h18004.www1.hp.com/products/quickspecs/13904_div/13904_div.html

3 new EVA P6350 dual controllers (available in HP EVA P6350 Combo or Starter Kits)

• EVA P6350 dual controller with eight 8Gb/s FC host ports
• EVA P6350 dual controller with four 8Gb/s FC host ports and eight 1Gb/s iSCSI host ports
• EVA P6350 dual controller with four 8Gb/s FC host ports and four 10Gb/s iSCSI/FCoE host ports

If you only have 8Gb/s FC host ports - you obviously cannot see any disks over iSCSI. 

As far as I can tell "Virtual Fiber Channel" connections in 2012 hyper-v only works with fibre channel, not iSCSI.

On the EVA you create a disk group, the physical disks are put inside the disk group.

From the disk group you create virtual disks.

These virtual disks you then present to hosts.

Hosts will see a block device.

If you want to split a vdisk up you need to do that on a server and then present the smaller portions on the server. Perhaps you can split this up with 2012 / Virtual FC?

You could also create a lot of vdisks. This can be scripted with SSSU (so you need to install Command View on a server with FC access).

With 25 disks you'll at least lose 3x900 for RAID parity (if system decides to use 8+8+9 RSS) and 2x900 if you use single drive protection. But if you run vraid0 perhaps you'll end up with more..

You want to make sure that all servers that access the same shared disk do this through the same controller.

[Drexel480]

Thanks for the info Johan but I can see that I left out some much needed information.

We are trying to decide between the:

• EVA P6350 dual controller with eight 8Gb/s FC host ports
• EVA P6350 dual controller with four 8Gb/s FC host ports and eight 1Gb/s iSCSI host ports

as controllers in the P6350.

So lets say we use the FC only controllers and do FC to the Hyper-V host machines and then have those Hosts installed with Windows 2012 for the Virtual FC support inside the Guest hosts.

Can I use Storage Works to create Vdisks on the SAN and then present them to the Guests hosts over FC or would the hosts only see Vdisks if they were using ISCI? Which would mean we would have to purchase the other controller with both FC/ISCI ports.

Or is the only way my FC Guest Hosts (VM's) would see the Volumns I want to present to them from the SAN is if we created Physical disk groups and chopped those up into small volumns?

The other alternative we were thinking was to go with the FC/ISCI controllers and just do FC to the main Host boxes and then ISCI for the Guest Host VM's. To me this would be slower but this will be a lab and we will not be pushing tons of data at one time.

But if we can do FC on everything that is our preferred method. We are just unclear on whether we can use Vdisk or not in this scenario. I personally am much more family with using ISCI since I have worked with Windows Storage server in small lab scenarios but now we want to grow our lab and put in a real SAN.

[Johan Guldmyr]

Hi, 

what you're asking is if the P6350 (and possibly SAN switches but also the HBAs)supports NPIV. 

The EVA supports NPIV. But does it support Virtual Fibre Channel(VFC) on the hyper-v 2012? Perhaps HP SPOCK has that answer.

iSCSI goes over TCP/IP. 

Because I don't think VFC supports iSCSI - I don't think you would be able to use VFC with vdisks presented from the iSCSI ports on a P6350. But perhaps hyper-v already has support for multipathing etc directly to hosts over iSCSI??

With controllers that do not have the iSCSI ports, you just present the vdisk(s) over FC to the hyper-v VMs.

thestoragearchitect had an introduction post a little while ago about vfc:

http://blog.thestoragearchitect.com/2012/10/05/windows-server-2012-windows-server-8%E2%80%B3-virtual-fibre-channel/

"Or is the only way my FC Guest Hosts (VM's) would see the Volumns I want to present to them from the SAN is if we created Physical disk groups and chopped those up into small volumns? "

So. On the EVA / P6000. You put all the disks in a Disk Group. From the Disk Group you carve out Virtual Disks (choose RAID and size for example). These Virtual Disks you present to hosts. Hosts see a virtual disks as a disk on the server. 

But I think you got it right. You'll need to create a lot of virtual disks inside the disk group (perhaps sorting with folders is a good idea).

"The other alternative we were thinking was to go with the FC/ISCI controllers and just do FC to the main Host boxes and then ISCI for the Guest Host VM's. To me this would be slower but this will be a lab and we will not be pushing tons of data at one time."

Meaning you would do iSCSI internally in the server - somehow use the Hyper-V hosts as an iSCSI target for  the Hyper-V Guests. 

If you do FC all the way, this seems like initially you'll have to create a lot of vdisks and present these to the correct vWWpN (virtual WWpN - that the Hyper-V host should assign to each Hyper-V guest FC HBA port). I'd say performance will be better if you can do FC all the way - no overhead while re-packaging packets into TCP/IP etc.

[Drexel480]

Thanks! That is exactly what I wanted to know! I really appreciate the help!