VMware Vxlan could be the catalyst for moving workloads across private and public clouds
Vmware Vxlan is an extension of Cisco OTV technology. It is essentially a tunneling protocol which encapsulates Layer2 across a Layer3 network, thus enabling flat L2 networks across data centers. It could prove to be an catalyst for moving workloads across datacenters or moving them to public clouds. The possibilites are endless as this enables new use cases like:
– Follow the sun
– Load balancing
– data protection in cloud.
1) The high level requirement is to retain the Layer3 IP address, so from a broad networking perspective VM
Mobility is transparent to high level networking services, initiators, targets and applications.
2) A consequence of this requirement is VMotion is only restricted to the same subnet, so when the VM moves
the VM’s MAC address and IP address remains the same, however the MAC address of the ESX server does change.
To mitigate this, RARP is used to notify L2 stack about the new MAC address. However, this retains the IP
address, thus making VM mobility seamless.
3) Since the subnet is same, there is no reason to use routing or Layer3, with the exception of long distance
Vmotion. Here Vmotion does work, but the IP address is not updated for the remote site (new subnet) and hence
it will not be seamless.
4) On dwelling a little further on this from a Long Distance VMotion perspective, there is a handicap in the
current validated design, which is, at the remote site, the subnet changes and hence the IP address needs to
change seamlessly. There are technologies Cisco is working on validating which employ Layer2 extensions and
Routing functions to make this seamless.