Most organizations in the cloud are struggling to keep up with their costs and usage as they relentlessly grow. Last week I discussed the basics of reserved instances as the first part of my blog series designed to help with the RI decision-making process. My second post in this series examines the cost benefits of using RI’s and how to use them across multiple AWS accounts.
For those new to AWS, it’s not uncommon to be hesitant to committing to 1 or 3-year reserved instances (RIs). The most common fear I hear is not wanting to purchase excess reservations for instance types that may not be optimal for your workloads in 6 months. In reality a 1-year term reservation will almost always break even after 6 months. This is when you can shut down an instance and still benefit from the reservation’s pricing discount. For a 3-year reservation, the break-even point usually occurs around 9 months.
Armed with this knowledge, the cost benefits become very tangible for organizations scaling their cloud infrastructure. Since the break-even point can vary by instance type, we use what’s called the payback period to calculate the exact number of months at 100% usage we need before we receive a price benefit. This metric is invaluable for mitigating the risks of reservations by identifying how long you must actually use them before they break-even.
The payback period is applied to partial and all upfront reservations, which are more expensive from a cash perspective earlier in their term due to the initial upfront payment to Amazon. It’s calculated by comparing the cash outlay for on-demand usage and the proposed offering over each month in a term, and then identifying the month at which the cost for the on-demand instance usage exceeds the cost for the reserved offering. There is no payback period for a no-upfront reservation, since they are less expensive than on-demand immediately.
Now that we’ve discussed the mechanics behind the typical break-even timeframe, let’s take a look at just how cost-effective reservations can be over time. In the example below, we see that an m3.large in the us-east-1a region will cost 37% less per month with a 1-year partial upfront reservation.
To calculate this, you simply need to use Amazon’s effective rate formula as shown below:
effective rate = upfront payment / reservation term / interval + recurring usage charges for interval
While the effective cost per month savings can be substantial, you may be wondering how reservations get applied across multiple accounts.
If you’ve purchased RIs in either the consolidated account or another one linked to the consolidated billing account, and there is no instance usage in a given hour in this purchase account to utilize the reservation, the reservation can be applied to matching instance usage in any other linked account within the consolidated bill.
Now this is where things can get tricky. Many organizations link one or more accounts together into a consolidated bill. Doing this unleashes one of the more powerful and, in some cases, confusing behaviors of RIs - their ability to “float” across accounts.
By default, reservations have an affinity for the account in which they were purchased, so the “float” will only occur if there is no instance usage within its account that can take advantage of the reservation. While the price reduction benefit of RIs “float” across accounts within a consolidated bill, it’s important to note that the capacity reservation does not. So if you have an available reservation in account A, Amazon will not guarantee that you can launch an equivalent instance in account B, even if these accounts are linked into the same consolidated billing account.
If you need help managing RI’s across your organization, see how CloudHealth can help you simplify the process of making your first RI reservation.
Stay tuned for our next blog post on the Life Cycle of Reserved Instances which will cover just how simple it is to model an RI purchase, the importance (and cost benefits) of modifying them on a regular basis, and, how to automate both of those processes.