iSCSI, NFS over the same VLAN, port, switch?

[acistmedical]

Our 2050 Setup per controller:

- FC to additional shelf

- FC to tape library

- 1G to CIFS

- 1G to iSCSI

- 2 available 10GbE ports (an expansion card)

So that's it.

We use 1 build in gig port for cifs access, one for iSCSI (for testing servers), which i want to move away from and use both ports for CIFS (2Gb per controller)

I will be implementing VMWare on blades and want to use NFS for datastores but still want to keep iSCSI for physical servers. I have 2 HP ProCurve 2910al 10Gb up 1Gb down switches that i will use. So i have 2 2 port 10Gb cards that i will uplink these switches.

So, can i use iSCSI and NFS on the same switch?

If so do i have to use 1 10Gb port on a card for NFS and another for iSCSI, or can i use both for both?

Then on a switch side, do i have to use separate VLANs and separate uplinks and switch ports per each protocol, or can both iSCSI and NFS use the same VLAN and then i will be able to use any port as either NFS or iSCSI?

BTW, These switches will not be used for anything else.

Thanks for reply.

[djaquays0]

Yes, you can use iSCSI and NFS over the same links, and I believe there is even a whitepaper *somewhere* on using iscsi and NFS over the same 10Gb links.  I wouldn't recommend it over 1Gb links for a production environment, though.

[radek_kubka · Accepted solution]

This is my take on this:

Ideally storage-related IP traffic should be separated from non-storage IP traffic - so for me having 'general' LAN & CIFS separated from iSCSI & NFS (for ESX provisioning), but iSCSI & NFS using same paths is good enough.

Having said that, there is the TR-3785 document showing feasible approach with just a couple of switches (although pricey Cisco Nexus ones) & couple of 10GbE ports per ESX host:

http://communities.netapp.com/thread/6276

Regards,
Radek

[acistmedical]

Thanks for reply Radek

This is exactly what i want to do, these 2 HP switches will carry only NFS and iSCSI traffic, CIFS and LAN will go through our Cisco infrastructure.

Thanks

Grzegorz