Getting ready for VXLAN

By | 2016-03-12

Background:

One of the breakthrough announcements made during VMworld this year was by Steve Herrod, the CTO for VMware, on the advent of VXLAN.  The way Steve Herrod described it, VXLAN was going to bring to virtual machines, the mobility that is currently afforded by mobile phone providers to mobile subscribers. Virtual machines would be assigned IP addresses.These addresses would allow them to be migrated to different clusters in the datacenter, without reconfiguration.

VXLAN is a proposed Internet Engineering Task Force (IETF) standard led by VMware in collaboration with Cisco, Arista, Broadcom, Citrix, and RedHat.  The draft standard can be found on the IETF website:

http://tools.ietf.org/html/draft-mahalingam-dutt-dcops-vxlan-00

The Technology:

VXLAN works by overlaying layer 2 networks over layer three networks.  What this really means is that the virtual machine’s Ethernet (layer two) network can be transported transparently over different TCP/IP (layer 3) networks.

Migrate between L3.vsd

The figure above illustrates this with two rows of datacenter racks that have different IP addresses blocks assigned to them.  In a non-VXLAN environment, the virtual machines would have IP addresses in either the 10.99.1-10.99.127 or the 10.100.1 – 10.100.127 blocks depending on the racks they were in.  This is usually because datacenters are usually designed with limited number of VLANS, and thus IP addresses, available in different locations of the datacenters.  This physical datacenter design would restrict vCenter cluster membership to ESXi servers in the same row.  This is due to the fact that the ESXi servers would need access to the same IP address used by the virtual machines they host.

VXLAN overlay networks break this restriction and make it possible for virtual machines to reside on any ESXi host. The virtual machines on those ESXi servers would be on their own IP subnets that would be carried over the physical networks available to the ESXi servers.

In Closing:

To get ready for all this flexibility in your VMware Infrastructure will not require you to overhaul your network gear, or make drastic changes in the way data moves around your network.  In fact the only change you might need to make, is to allow for multicast on the network segments you want to carry VXLAN traffic.  Support for VXLAN will be included into the next version of the virtual distributed switch, as well as the upcoming release of the Cisco Nexus 1000v virtual switch.

As you can see VXLAN will allow for greater flexibility in how you design your VMware Infrastructure and provide for easier mobility of virtual machines in and around your datacenter.

Leave a Reply