Solving the VM Mobility problem across multiple Data Centers at Layer3

July 19, 2013 Markus Nispel

We are on a roll here with our Data Center Interconnect solution that has been a key component of our latest Data Center launch. The market – press, analysts and key industry experts – cheered after we explained them how easy it can be: the beauty of simplicity that I focused on in one of my blogs before the launch took place. Just released the layer 2 fabric extension with both fabric and host routing to optimize east/west and north/sound traffic in a active/active multiple data center mesh we are taking this to the next level soon: enable VM mobility even in the absence of a layer 2 extension technology (IEEE SPB or LACP via Virtual Private Port and Ethernet Services, GRE/L2 based) that we just introduced. The picture below shows this concept of a layer extension with a stretched subnet where we announce host routes for VM A and B so always the best path is chosen. As those VM move the host route moves with them.

If there is no requirement for Layer 2 Protocols to the VM then we can enable VM mobility, without complex tunneling techniques like VXLAN and NVGRE that the hypervisor switches promote, across multiple fabrics. Sounds easy – and it is: enhancing our IP host mobility (host routing) solution to work also in foreign subnets a VM can be moved into any data center and any subnet. Leveraging the fact that the ARP cache entry in the VM is identical for the master VRRP router when the same VRID is used we can move the VM without any constraints while providing continuous connectivity after the host route got propagated from the new nearest (nearest to the VM) router. Sounds easy? Yes it is – again, the beauty of simplicity. As with OneFabric this is not just a term we throw out there, we design our solutions with those principles in mind.

In this example the VM B, which has an IP address in the same subnet range as VM A got moved using vMotion techniques. To maintain connectivity the router in the foreign subnet propagates the host route of VM B after it detected its local presence. So other VM and also clients can reach VM B on a optimized path.

The only thing we have to do now: spread the word! Simplicity makes the difference.


The post Solving the VM Mobility problem across multiple Data Centers at Layer3 appeared first on Extreme Networks.

Previous Article
The network as a business asset – towards analytics and business intelligence

In my last blog about application visibility and control I did mention analytics as one of the most interes...

Next Article
What Keeps a Technology Executive Up at Night

Ever wonder what legitimate concerns keep tech executives up at night, or what is just hype? Check out the ...