AWS Reserved Instances come in a bewildering variety of shapes and sizes. Buying and managing them is no easy feat. In an effort to tame the complexity, Amazon recently rolled out Savings Plans. These represent a huge change to how AWS reservations are bought and managed.
They work by requiring the customer to commit to a certain level of resource usage, measured in dollars per hour, for a one year or three year term. All resources under that limit are covered by the plan, while usage over the limit gets charged expensive on-demand prices.
Overall Savings Plans are an intriguing new option, but they’re not a one-size-fits-all solution. Our analysis shows that carefully combining Savings Plans with existing Reservation offerings on average results in an additional 5% savings on your total AWS bill.
Even if you decide to go with a pure Savings Plan strategy, combining multiple Savings Plans with strategically crafted commitment sizes yields better savings than a single one-size-fits-all plan.
Digging into the Data
As a team of ex-AWS data scientists, we know there is always a gap between AWS marketing of a new savings product and the actual savings data from customers trying to use it. We did a detailed analysis to understand the impact and pitfalls of this new savings vehicle for you as a customer.
Savings Plans come in two flavors, Compute and EC2 Instance. Roughly speaking, Compute Savings Plans are analogous to Convertible RIs (CRIs), and EC2 Instance Savings Plans resemble Standard RIs. Compute Savings Plans/CRIs offer more flexibility while EC2 Instance Savings Plans/Standard RIs have higher savings.
The numbers show that CRIs are almost strictly worse than Compute Savings Plans. The savings plans have the exact same savings rates as CRIs but offer additional flexibility and require no management. Like CRIs, Compute Savings Plans still need to be mixed and matched with different upfront commitments and term lengths to get the best deal given your constraints.
When we look at the less flexible options, the trend switches. In this case our data show that the old Standard RIs are often a better deal than EC2 Instance Savings Plans. Instance Plans are useful if you are changing operating systems or tenancy frequently, but this isn’t a big concern for most customers.
The locked in nature of Savings Plans punishes you for fluctuations in your hourly resource usage for a certain EC2 machine family. If you commit to more than you need, then every hour you use less than your commit level, you’re wasting some of the value you paid for. On the other hand, if you under-commit and consistently go over, you’ll end up paying for on-demand - the highest pricing tier. This puts purchasers in something of a Catch-22.
Reserved.ai solves this problem by replacing locked-in Savings Plans with carefully managed Standard EC2 RIs. Standard RIs can be bought for shorter terms than one year and if your needs change, you can sell them back. In fact, we guarantee and automate the buyback of under-utilized Standard RIs. We also keep an eye on your usage and alert you to unexpected changes in reservation coverage.
The bottom line is, while Savings Plans are a welcome new option, they won’t optimize your spend all by themselves.
Optimizing your costs
What you’d really like is a plan that leverages all of the AWS pricing options to optimize for your unique needs and priorities, but this seems to put you right back where you started: trying to make sense of a complex array of pricing schemes. Sorting through it probably means assigning team members to spend lots of time trying to make sense of AWS pricing arcana.
This is where Reserved.ai comes in. We let you set whatever constraints you want in terms of upfront dollars and commitment and find the perfect mix of RIs, Savings Plans, and other resources to get you the coverage you need for the lowest possible cost. Using a plan tailored to your needs has two primary benefits.
- More savings: It’s often possible, by using the right mix of different purchasing options, to get you the exact same resources for a lower price.
- More flexibility: You’re not limited to a single one year or three year contract, or upfront spend of 0%, 50%, or 100%.
You control everything from one screen and don’t have to worry about the fiddly details. This frees up your team to focus on more important projects for your company.
This table summarizes the average savings our customers see with Reserved.ai purchase recommendations compared to Savings Plans recommendations through AWS Cost-Explorer. As you can see, we’re able to offer incremental savings regardless of your preferences for length of contract or upfront cash. How do we do this?
While Savings Plans lock everything into a single contract, we can mix and match various resources to get you the best deal. For example, we might find that a mix of Compute Savings Plans, Standard RIs, and Database RIs yields the highest savings. You enjoy the benefits without having to deal with the complexity - it all happens behind the scenes.
Combining reservation purchasing across all reservable AWS services allows us to leverage your upfront dollars more efficiently. We automatically apply them to the areas where they produce the highest incremental savings.
In addition to greater savings, we also offer more flexibility, both in terms of initial options and making adjustments later. While the table just shows the direct comparison of our mixed options to AWS Savings Plan recommendations, we actually allow you to choose any dollar amount you want for upfront commitment - not just 0%, 50%, or 100%.
All in all, Savings Plans are an attractive new option, but they’re not the end-all be-all of purchasing options. Getting the maximum possible savings often requires a mix of Savings Plans, Reserved Instances, and other resources. Reserved.ai works to get our customers the best deal and highest savings rates on their cloud commitments. Savings plan are another arrow in our quiver to get you the lowest possible price.