So on the one hand we have makers of big servers proposing virtualization as a way of "optimising the use of powerful and expensive assets", and on the other hand we have makers of blades (often the same companies) promoting solutions to combine those blades into clusters and grids to get maximum computing power and redundancy.

Can I suggest a new idea? If you want a f***king big system, buy one. If you want a hundred servers, buy a boxful of blades. Why buy one, and then spend a lot of time and money to make it look like the other?? Can anyone actually explain this to me?

I agree a lot with this article and that virtualisation is really old hat. Difference is that it is available on commodity hardware.

With all systems there is a small difference say 20% which gives the server hardware leverage and a competitive edge. But this 20% of expensive but extra reliability due to special h/w e.g. memory mirroring and things like dual internal crossbars, gives 80% of the value. A badly managed mainframe with unskilled operators has availability equivalent of a desktop PC. A high end unix server now has all the RAS features of a mainframe e.g. RAM memory mirroring and instruction retry. But you need good people to manage those high end unix boxes. Mainframes and Unix servers have been virtualisating/consolidating small systems for a long time. Doing what VMware does for a long time. Are not the VMware developers old unix/mainframe people. The equation is hw + people skills + virtualisation s/w.

Also, as all old timers know, virtualisation has a cost, the overhead of the virtualisation layer. We should not forget this.

So it is the quality of the support staff which make systems reliable. Outsources manage this as they try to never touch/change a system. The trick is to be able to change/upgrade systems and add new apps while maintaining availability. Datacenter grade Unix servers have allow hot swap and dynamic changes on systems which is ideal for a changing virtualisation environment also hardware to cope with failure of chips, memory, i/O etc. So very little risk here. I do not see this available yet on commodity, why we do not have it on X86, these extra features cost money. You get what you pay for.

So bleeding edge is virtualisation on X86, but put all your print servers on one X64 box and you have a single point of failure e.g. risk. Can't print anymore, save the trees, this is green IT. Virtualise with commodity low paid, unskilled staff on commodity hardware and you are treading a fine line. All eggs in one basket with people not used to carrying eggs.

How long will this fashion continue, I reckon on another 18 months, virtualisation is already about 18 months into it's cycle. Most IT fashions last for about 3yrs.

What is the next fashion, easy, black. Black is the new black.

What will the future be like, simple it will be different.

Perhaps Dale will clarify: what I find interesting about consolidation and virtualisation today is the continuum of degrees of separation which can now be achieved:

Resource Management (most Unices)

Zones/Jails/Containers

Hypervisor based hardware virtualisation

Hard domaining

It is particularly the middle two which are of interests: the former allows separate "virtual systems" within a single OS instance, while the latter in its various guises allows commodity hardware to be sliced up for multiple virtual hosts.

Please, update your information !! Novacaine is so yesterday. Lanacaine at the very least. Even my butcher of a fangman used Lanacaine to extract some teeth (and the subsequent extraction of a hefty wad from the wallet/credit card) !! The dentistry may be painless but the subsequent extraction wasn't !!