I’m going to hazard a guess here and think that its likely that at least 90% of people out there would not consider a lovely sandy beach as a solid foundation upon which to build a home, regardless of how nice the view is. In order to build that house you’d need to have a survey done on the bedrock, if any, and then build an actual foundation which the house would then be built on. Build the house on sand and it will sink.  I think you can search Google News for hits on disasters in Malibu and Laguna Beach, two very swanky neighborhoods.

VMware had the same idea, apparently, or at least they picked up on the ideas of their incredibly smart customers and turned them into products. The reason I’m talking about this is because I am in the process of virtualizing an entire operation. Everything, even Exchange 2007 which is about to be tested for a few weeks on a sandbox HP GL360-G6 (damn nice server, kids!), is getting the treatment, and we’re standardizing on VMware’s ESXi.

One of the more interesting aspects of this virtualization project is the fact that we are extending the concept to single machines. Nobody with experience would install Exchange on a box which was also the PDC, BDC, print server, file server, web server, and whatever else (oh! I’m talking about SBS, aren’t I!! but that’s another editorial). The best practice is to install it on its own dedicated server, so why not just put ESXi on that server first? Sounds weird, doesn’t it? Why would you virtualize a server host if you’re only going to put one guest OS on it? I know it sounds radical, but it makes sense and isn’t complex to understand. What do you get when you install an OS into a virtual machine container? You get a single set of predictable hardware in a self-contained, portable, replicable computer. As long as you have a VT-x capable host and can install a copy of ESXi, you can make that machine run. In an enterprise NOC environment, this is fantastic. You don’t have to be as stringent in your hardware purchases in order to maintain homogeneity. After all, every sysadmin knows that homogeneous configurations are the easiest to maintain and keep up.

In our NOC, we are focusing on disaster recovery without the concept of baremetal restores. All of our hosts are ESXi (we’re installing vSphere and vCenter soon) and we are running everything from Windows 2003 R2 and Windows 2008 R2 to Ubuntu Server. We also have a 4TB NAS in place (dual gigabit) which is allowing us to test running VMs from it instead of using local storage on the host. That’s working out quite well, though I’m wondering where the performance cap is. There must be a point at which the bandwidth required for the VMs to pass data will saturate those two gigabit connections and start to degrade system performance.

Looks like I need to point WireShark at the NAS to take a peek at the kind of traffic its passing. A little bit of maths after that and I should have an idea what our upper limit is. More on this later on and I update the Chronicles!

Purchase this and related items at Amazon now: