Recently I gave a talk on using Cisco UCS servers with a Virtual Desktop Infrastructure solution. One of the points I made was about the statelessness of Cisco UCS. What exactly do we mean by that? Essentially we’re talking about Service Profiles here. Service profiles are essentially the personality of your server. You can create a service profile by creating policies for things like VLAN pools, number of NICs each server gets, number of HBAs, and whether the server will be boot from SAN. The boot from SAN element is important in stateless computing. Once you finish making the service profile you can associate it with a physical server, either a blade or a rackmount UCS server that’s being managed by the UCS Manager.
Now, imagine that you have a server failure. All of your virtual desktops failover through the use of HA to another server, but you’re probably running on less resources than you’d like. At this point you can disassociate the service profile with the failed server, pull it out and put a new server. Once you’ve put that server in you can associate the same service profile with the new server, boot from SAN, and within minutes that server is up and running and looking exactly like the old server. Your virtual desktops can fail back and your environment is running as it should be.