Nexus 1000V connected to multiple networks

[thilo schueler-mach]

hello,

our VMware Server are connected to multiple networks. To each network we use port-channels for load-balancing and availability and we use 802.1q links to each customer/network. Also sometimes different customers use same VLAN-ID`s and we could not avoid this.

With VMware vNDS we have no problems with this requierment, because we can create seperate vNetwork Distributed Switches to each network/customer.

But Nexus1000V knows only one Switch per Host, so we can not seperate identical VLAN-ID´s from different customers.

I read some documents about Pinning, but this Feature is only for load-balancing with port-channels and not for selection of Uplinks.

Does anybody knows, if Cisco will implement switching instances in Nexus 1000V or Q-in-Q ( double tagging ) ??

Both implementations would resolve our problem.

thanks,

Thilo

[lwatta]

This has also come up a few times. We are working on a solution to allow you to have the same VLAN but different paths. It should be possible when we release Virtual Security Gateway to provide a solution but as far as a native solution with just Nexus 1000V we are working on it.

louis

[artm.reg1]

Hi.

Is there a solution for this now?

[thilo schueler-mach]

I believe the solution is called Network Segmentaion along with VXLAN, vShield Manager and vCloud Director. Maybe an implementation like vDS where you could provide the same VLAN-ID on different physical interfaces is just to easy. 

[artm.reg1]

Sounds like a complicated solution for an easy problem.

Sadly, I will have to stick with VDS for now.

Are there any plans to support multiple VEM per host in the future?

Thanks!

[thilo schueler-mach]

Much more easier and standardised could be a solution with Q-in-Q. With a support of multiple S-VLAN´s per interface it should be possible to provide the services to multiple customers also via one interface. Then a customer aggregation switch could push and pop the outer vlan tag.

Regarding your question about multiple VEM, until now I have never heard about it. And maybe there is a disadvantage with the scaling limit of 64 VEM per VSM.

Any ideas why Cisco doesn´t provide Q-in-Q on N1K?

Thanks,

Thilo

[artm.reg1]

Is it possible to assign a different vxlan for each network & still use VLANs inside the vxlan?

how would I connect the non-vxlan supported physical network in layer 2?

Can I make an ethernet uplink port-group part of a vxlan & still use it as a vlan trunk?

Thanks!