EC2 Reserved Instance - Are 24 instances running 1 hour (in the same hour) equal to 1 instance running 24 hours?

Question

I understand that reserved instances in AWS are more of a billing thing rather actual instances - they're not attached to an actual instance - and I wonder:

If I buy a reserved instance of a specific time in a specific region and availability zone - will I be billed the same rate if I use a single instance 24 hours a day vs. running 24 instances (same size, same availability zone) for 1 hour (same hour, e.g. 00:00-01:00) every day?

Is there a difference in the types of reserved instances with regards to this question?

If the answer is no, and obviously the case is not such a clear cut (i.e. I have instances coming and going all the time - some work simultaneously and some on different times) - how can I make the calculation whether or not it makes sense for me to purchase reserved instances?

Answer 1

I believe not.

It is not obvious from their product description, but it stands to reason, that "reserving an instance" and "long term commitment" should not be abused to have 365 instances running 1 day be as cost-effective as having the intended 1 instance for a whole year.

Using this logic you would also ask - why should I reserve more than one instance? I'll just bill all my instances on a single instance commitment - hardly a commitment, right?

Anyway, on my billing (where I have some reserved instances) I see:

USD 0.1905126763 per Linux/UNIX, m1.large instance-hour (or partial hour) (blended price)*

I believe that blended price mean that I have more instances than I reserved, so the hourly fee is a blend between the regular hourly fee ($0.24) and the reserved instance fee ($0.084)

Update

From AWS's documentation on Blended rates (thanks @user3319714):

Linked Account 1 has purchased three Reserved Instances and has used 2,100 hours of the reservation. Due to fluctuations in application load, 60 reserved instance hours remain, which can be applied to other eligible usage in the account family. In addition, application load when all three reserved instances were already running has necessitated an additional 40 hours of on-demand usage.

From this it is obvious that running more instances that you've actually reserved will not be billed are reserved.

Answer 2

Certainly Not

Reserved instances are a mean to provision the required infrastructure to handle the demand, and are cost effective in the long term.

In the presented scenario you would be billed for processing time of 1 reserved instance and 23 onDemand instances.

You may try to find a balance between using reserved instances, onDemand instances and spot instances.

For detailed information take a look at the following documentation.