> EC2 works for a lot of models, but one that it does not work for is
> small traffic apps that need to be available 24/7. If you have a small
> instance (AWS term) running full time with a fixed IP, it costs about
> $75 a month. If you turn it on for 2 hours a day, it costs about
> $15/month. A large instance is about $325.

If you intend to run anything 24/7 for more than a couple months, a
reserved instance is the only way to go as it will drastically reduce costs
-- for example the small instance you mention above should be a bit above
$25/mo including the reserved instance fee if you can commit for awhile.

If you do a lot of experimentation, you can buy a reserved instance and
create and destroy servers at will while paying really low rates (0.02/hr
is the rack rate for small reserved instances). If you can build things to
deal with a bit of uncertainty, you can pay rates like that without even
reserving the instance buy purchasing CPU on the spot market.

The micros are a bargain for low intensity apps that don't require much
horsepower (e.g. proxy, backup, etc). A reserved instance goes for $0.005
per hour (roughly the same price on the spot market), and even full hourly
rate is 0.02.

One thing to be aware is that many organizations effectively shift
substantial costs such as bandwidth onto other entities. Amazon is unlikely
to line out on your balance sheets favorably if your application puts
substantial demands on a resource that is cost shifted. Amazon is also less
attractive if you have a need to manage a low value data that doesn't get
much use on active disks.

There are other providers, and it's really worth looking at what your
actual needs are before deciding on an IAAS provider as the cost structures
vary quite a bit. We looked at others, but Amazon was a better fit (and
cheaper for meeting our specific needs).