[Someone1234]

Uptime guarantees are rarely worth the paper they're printed on. Most will give you some lame credit if they exceed the very generous amount of downtime they allow (e.g. 99% = just under 4 days a year downtime).

The only way to get real uptime is to have dedicated on-site support. Which is to say "someone to yell at" when the system(s) go down. But that type of service is quite expensive indeed (add an extra 0 to what you're currently paying minimum).

For $100 you're going to get shoddy support and some downtime. All you can really do is buy $100 systems with different vendors so that if one vendor is offline you can push the workload to the other two.

But then of course you need to design your infrastructure to move workloads around, and that within itself is a lot of work...

[rcorin]

Agreed, we do have several vendors and do distribute the workload as much as possible. In our experience when a server goes down the whole data center has problems, so even if it's more inconvenient the best is to really have an assortment of vendors.