cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
974
Views
0
Helpful
6
Replies

UCS basic concepts

ciscoworlds
Enthusiast
Enthusiast

I'm working on DC/Sec technologies. would you mind, please confirming me on this? 

1. suppose we're using 1240 card on a B-series server and have created a service profile with 4*vNIC and 2*vHBA. As I know this mezzanine card lets us to create 256 virtual card (including vHBAs and vNICs). so if we assign this service profile to this B-series server, does it mean that this server just has 4 LAN interface and uses them to connect to the rest of the network plus 2 of HBA card to connect to the SAN? does it mean that we now have 256-6=250 VIC remaining ready to be assigned to the VMs that will be created on the same B-series server in the future?
supposing that all of these assumptions were correct, does that mean traffic of all of the VMS that will be created on the same B-series server eventually have to exit off the same physical B-series server through 4*vNIC interfaces that were created before as result of service profile?
2. links between FI and b-series chassis are special links that carry VN-tagged traffic. the only way of using FCoE in the UCS environment is configuring "uplinks FCoE" ports on the FI, because FCoE traffic can internally handled up to the FI and we don't need any speciall FCoE-related config to be done on the servers. that means the only available FCoE config on UCSM is "Configure as FCoE Uplink". am I right? with this in mind, if we had 3rd party server with FCoE card on it, the only way to connect FI to it, is setting up the port on FI as "FCoE storage port" and changing the FI mode to FC switching mode. am I right?
1 Accepted Solution

Accepted Solutions

Greetings.

Unless you are using VM-Fex (for ethernet) , or  vHBAs at the guest VM level(this is only supported on hyper-V currently),  the vnics and vhbas defined at the service profile level are obscured at the hypervisor level, and the storage seen by the VMware guest VM is just a .VMDK file parked on a datastore the ESXi host has mounted.

Your upstream device configuration/capabilities will determine whether you setup FCOE or FC uplinks on the FIs.

Thanks,

Kirk...

View solution in original post

6 Replies 6

Kirk J
Cisco Employee
Cisco Employee

Please see http://www.cisco.com/c/en/us/td/docs/unified_computing/ucs/ucs-manager/Reference-Docs/Configuration-Limits/3-1/b_UCS_Configuration_Limits_3_1.html#reference_6A3C0158393549E5A3E3F87002234145 for vnic/vhba maximums.

These are actually governed by the OS limitiations.  You hit the OS limitiations long before you get close the the UCSM/hardware limits.

If you are going to attach FC/FCOE storage processors directly to the FIs then you will need FC switching mode as that will allow you to do the required local zoning.

If you intend to connect to FC/FCOE storage that is northbound, then you will want to setup the appropriate FC/FCOE uplinks, along with the corresponding VSAN and underlying FCOE vlan.

You will reference the vsan on the vHBA in the service profile, and on the FC uplinks.

Connecting Non Cisco servers to the FIs is not supported.  Exception is an appliance storage port  >  IP storage such as NFS or iSCSI

Thanks,

Kirk...

Hi and thanks for your reply. It's better to ask my question in a very simple way; supposing I'm going to install ESXi on a b-series server, is it possible to configure FCoE on that ESXi server, so that server uses FCoE on it to connect to both Lan/San networks?

Greetings.

While FCOE is the transport used within the UCSM, the bare-metal OSs simply see a regular FC HBA, and you do not need to configure FCOE at the OS level.

If your FIs have native FC ports connecting to upstream SAN switches then the Ethernet encapsulation is removed and the FC frames continue northbound.

It's the flexibility of the VIC adapters & service profiles that allow you to create both Ethernet and FC hbas that are presented to the PCI bus of the Host. 

Thanks,

Kirk...

Thanks a lot Kirk. based on what I understood from your comment, the bare-metal OSs and even VMs inside that bare-metal OS just use vNIC (for ethernet) and vHBA (for SAN), then traffic handled "internally" and "automatically" by UCS up to the FI without our interactions (regarding to the FCoE). Only from that point up to the network core is where we can use either of native FC/Ethernet or FCoE. would you confirm me on this? tnx again. 

Greetings.

Unless you are using VM-Fex (for ethernet) , or  vHBAs at the guest VM level(this is only supported on hyper-V currently),  the vnics and vhbas defined at the service profile level are obscured at the hypervisor level, and the storage seen by the VMware guest VM is just a .VMDK file parked on a datastore the ESXi host has mounted.

Your upstream device configuration/capabilities will determine whether you setup FCOE or FC uplinks on the FIs.

Thanks,

Kirk...

so if we assign this service profile to this B-series server, does it mean that this server just has 4 LAN interface and uses them to connect to the rest of the network plus 2 of HBA card to connect to the SAN?

Yes.

does it mean that we now have 256-6=250 VIC remaining ready to be assigned to the VMs that will be created on the same B-series server in the future?

No. You assign to the VMs virtual ports of vSphere virtual switch.

does that mean traffic of all of the VMS that will be created on the same B-series server eventually have to exit off the same physical B-series server through 4*vNIC interfaces that were created before as result of service profile?

Yes. If all of vNICs assigned to the same virtual switch. But you may create 2 virtual switchs with 2 uplink each.

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: