Discussion on:

Are operational considerations overlooked in virtualization?

Excellent article. I keep on seeing organizations that seem to think that Virtualization will solve all their problems. But your absolutley right that you need to get the requirements down first before you go virtual and see if it makes sence for both cost and operations. Ive seen companies spend more money going virtual than what it would have cost by having individual servers because of they thought they were saving money.

Sadly, I've seen virtualization used to add more servers on less hardware with horrid results. They don't have enough money for hardware, so they just add another virtual server to an already overly full set of virtual servers. If you are running servers that require a lot of networking traffic, add more physical network adapters rather than make 8 virtual servers share one nic. You also need better redundancy so when hardware fails (which is known to happen) you don't lose all 8 servers for a week.

In our case, virtulization did cost us more than if we bought individual servers, but this left us with capacity to add more servers. This was done intentionally, as we knew what some of our future requirements were going to be and in the long will save us money.

We have also set up redundant hardware as well as virtual servers, if anything goes down, the other takes over instantly until we can get everything back up and running properly.

There are two physical virtual machines, our storage has redundancy and there are two of each virtual server, we have redundant network switches, bottom line is everything has redundancies.

Yes, this is costly, but look at how much downtime costs a company, even a small business could have a cost of $5,000 per hour of downtime and that is just in overhead costs, not talking about lost revenue.

I'm a backups guy working exclusively with Tivoli Storage Manager from IBM since 1998.
Recently I've seen multiple large shops get the Virtualization Religion and virtualize everything without adequate consideration of each apps real requirements. They particularly hate it when I tell them that the TSM backup server has to be on physical hardware.

Mistakes I've seen are - VMing multiple SQL Server machines with one instance and a couple of databases each rather than running a big iron 3+1 cluster with multiple instances and more databases in each instance - and - virtualizing multiple Notes machines rather than running several Notes instances over a couple of real machines with app clusterimg.

In these cases the big iron will out perform the VM solution handsomely and with much the same availability. As always in IT you need to look at the pros and cons for each app separately rather than jumping on the latest bandwagon.