Nexus 7000 vPC+ design/configuration

[Marius Gunnerud]

Hi,

We are in the process of planning and designing our new network.

We have a few servers that have bundled interfaces and will be connected to two FEXs which in turn are connected to two Nexus 7009 switches.  These two N7K switches are connected to another remote data center via Fabric Path with redundent links between them.  The issue is that these servers need to utilize only one its interfaces because the software which is being run on it requires this.

The two data center locations are required to be seen as one logical data center.

I am wondering if it is possible to use vPC+ between the two data center locations to configure a makeshift HSRP setup making the N7K switch connected to the "active" server interface a sort of active switch while the other is the standby?  If this is possible, could you provide a link to documentation that describes this?

Thanks,

Marius

[sean_evershed]

Hi,

I'm not clear, you have vPC+ configured between the two data centres today or you plan to use it in the future?

The Cisco Validated Designs recommend either OTV or VPLS as the primary choices for Data Centre Interconnectivty (DCI).

http://www.cisco.com/en/US/netsol/ns749/networking_solutions_sub_program_home.html

It is also recommended to localise HSRP to a single data centre and to not allow it to communicate with another data centre. There are several presentations at Cisco Live on DCI and vPC+, for example see the session BRKDCT-3060 from Orlando 2013 on Deployment Challenges with Interconnecting Data Centres.

Don't forget to rate all helpful posts.

[Marius Gunnerud]

Hi Sean,

Thanks for your reply.

We are currently in the design phase and none of this is currently configured.

I  understand that Cisco recommend OTV or VPLS, but from my understanding  that is because most companies can not afford a dedicated dark fiber  between the site. We do have a dark fiber between the data centers and  therefore fabric path is more viable choice for us.

Also,  OTV, again from my understanding, produces more overhead than we can  afford to have.  so therefore again, we see fabric path as the obviouse  choice.  Please correct me if I am wrong about OTV producing overhead.

But we have the small dilemma with some of the servers  needing to send and receive data on the same interface eventhough they  have bundled interfaces.  so we are looking into how to get traffic to  prefer a defined switch which is connected to the interface that we  would like the server to use.  Do you know if it is possible to  configure the switches so that the other datacenter will prefer to send  traffic to one switch rather than the second switch?

Thanks agian