I guess that most of you have heard about the changes in the Amazon AWS Reserved Instances offering.
Besides the Convertible Reserved Instances, that I may cover later on, I was a bit baffled with the "Regional Benefit" concept and what really meant to "waive the capacity reservation" to achieve it. So I asked our account manager, and thought useful to share his comments here.
- The new default (in fact, what you get when you click on "Purchase more like this" on any existing Reserved Instance in your account) is the "Regional Benefit" one.
- With reserved capacity, you make sure that if at any given point in time in the future (while the RI is active) you want to launch an instance of the specific instance type of the reservation and in the specific Availability Zone of the reservation, you will be able to do so.
- Without reserved capacity (so if your RI has the "Regional Benefit"), your instance (again of the specific type of the reservation) may launch in any Availability Zone in the region where there is capacity or (very extraordinary circumstances) not launch at all, if there is no additional capacity in the region.
- This may only happen e.g. if a large customer launches plenty of instances of a given type in a given AZ in a region. This can lead to a situation where this instance type is temporarily unavailable in a given AZ or a whole region. It has happened in the past, but very rarely (as per his knowledge).
- Generally with most common instances (t2.small, m4.large ...) this should never be a problem.
In our case, we are buying Reserved Instances for already running servers, so the "Regional Benefit" benefit is very obvious (we don't have to worry in which AZ the instances are running anymore, which is a big relief).
I can't really think of many use cases for sticking with reserved capacity Reserved Instances.