>We decided for own HW as our bottleneck was disk troughput: we have dedicated SSDs for a couple of VM's where there is dire need.
>In case of HW failure those VM can either run on another dedicated server in the same rack or on a setup similar to proposition #2 if older backup HW cannot cope with the emergency - price agreed upon before, but will only be billed if that HW failure occurs and vendor HW is used.
>
>Look at your own NEEDS - where are you bottlenecks ?

We are considering more into the public cloud hosting now. I do not like the fact this would bring too much new hardware in a new solution. Public cloud hosting have very big infrastruture already tested and everything is virtual. So, I can be set up in two weeks and control the resources as I need them. For dedicated hardware, everything is physical and when more RAM would be needed, for example, we would have to rely on support ticket, install new RAM. This would also not offer unitary RAM or Core expansion. With the other one, I can decide to add 1 GB of RAM and 2 Cores and I would pay an extra 1x8.92$ + 2x15.00$ on the next month. The same would go with the disk space and the required bandwidth.

For our needs, it appear this would do. Our applications are not file based oriented but only database. So, that would eliminate a lot of disk I/O.

The monthly cost of dedicated harward would be about 1300.00$ to 1500.00$ in difference. So, that would be about 18000.00$ more per year. That solution also requires 6 week in advance as they have to purchase the hardware, install it, etc.