FCoE VLAN and FC Uplinks in Cisco UCS Architecture
Posted by craig 
- on March 10th, 2010 in Data Center, Server, Tips | 6 Comments »
Today I found some interesting technical details during my installation of Cisco UCS for the client. FCoE architecture are slightly different when you decide to uplink the FCoE traffic end to end to the Cisco Nexus 5000 unified fabric, or you are doing the FC uplink directly from the UCS 6120 to the MDS or SAN fabric switches. I went through the documentation and my previous round of installation, we were actually configure the LAN & SAN uplink from UCS 6120 to Nexus 5k. Therefore, in our configuration, we required to have specified FCoE VLAN which carry the FCoE traffic to reach the storage through Nexus 5000. Today, I implemented the similar configuration, but the different was the uplink configuration.
This new setup are included with 10 G Ethernet uplink from UCS 6120 to Catalyst 6509 through Fiber, and utilizing FC Uplinks from UCS 6120 to MDS 9124. In my test, I removed the VLAN that I created purposely for FCoE, and found that there are no impact to the configuration. I had confirmed with the support guys that if you are doing an uplink with FC from UCS 6100 to MDS, you will not require to tag the specified VLAN that created for the VSAN in the configuration.
6 Responses
UCS has default behavior, it will create a VLAN with same ID as VSAN.
But I forgot where I got this.
Have you used M81KR. A very good adapter. support more vNIC and hardware VN-LINK, it can replace nexus1000v VSM.
Had not tried the M81KR as most of the UCS deployment I recently did are more for virtualization. As my understanding, the M81KR will virtualize the 10G to multiple vnics to achieve separation in some scenario the system require multiple interface with Gigabits performance. This will not replace the Nexus 1000V. The replacement for Nexus 1000V is Nexus 1010 which is applicant base that come as a blade which are compatible with the UCS chassis.
I have personally used both the Nexus 1000V and also UCS. I can state for a fact, that except in niche scenarios, the Nexus 1000V is no longer required. The fabric interconnects effectively become the VSM and connect with vCenter to manage VM virtual circuits.
Hi Lee, I assume you are referring to the blade server provide the VSM functionality, correct me if I am wrong.
No, nothing really to do with any blade server. The VSM functionality is provided by the UCSM. The vDS is set up and managed through the UCS-vCenter integration. As with the 1000v, each ESX host still runs a VEM.