Ask HN: Can you explain Amazon EC2 Reserved Instances?

I am trying to price up running an instance for 12 months. Part of that is trying to understand their reserved instances and if savings can be made by buying them.

Unfortunately Amazon's pages are cryptic at the best of times and damn right nonsense in some cases (or maybe I'm just a moron).

So I have the following questions:

- If I get a 12 months reserved instance, do I need to buy a "on-demand" instance first? Are you reserving the on-demand instance or is a "reserve" instance a type of instance within its self?

- What the heck is the "Offering Type" (or RIs) and why would a Heavy cost LESS than a Light instance? Shouldn't it be the other way around?

Just to give you an example. Let's say we want a "medium" Linux instance for 12 months. 100% monthly utilisation.

If we do a single on-demand instance that is $117.12/month. But if we do a single reserved instance that is ONLY $23.42/month.

That seems "too good to be true" in terms of discount/savings (80% off). So I am guessing that isn't how Reserved Instances work.