Since Savings Plans were launched by AWS to help cloud consumers reduce hourly rates via committing to ongoing levels of spend, they’ve become hugely popular. This is no surprise given how flexible this instrument is and how it simplifies the purchasing process compared to Reserved Instances. On the flip side, it does become more difficult to understand usage that qualifies for commitments, predict future patterns, and assess the likely impact of additional purchases.
New Savings Plans recommendations from Apptio Cloudability, a solution for optimizing cloud resources for speed, cost, and quality, give you the visibility and insights to reach your commitment discounting goals.
How commitment-based discounting at AWS has evolved
The original instrument, Reserved Instances (RIs), allowed users to commit to specific types of usage for particular cloud services for either one- or three-year terms and receive a discount of up to 72% in return. For example, you could purchase a one-year, partial-upfront m5.large Linux EC2 Standard RI in the us-east-1 region and receive a 40% discount over on-demand prices. That’s of course presuming that in every hour of that 12-month period there was an EC2 instance running somewhere in the organization that qualified for that benefit. With a cloud footprint of any scale, matching usage to possible RIs becomes complex, and it’s the reason Apptio has always advised to base your RI program around understanding individual breakeven points and setting a waterline.
The introduction of Savings Plans changed this landscape by compressing the classifications (combination of instance family/size, operating system, and region) that you commit into and the purchase unit (dollars instead of instances). Over time, it’s also opened commitment-based discounting to more services like Lambda and Fargate, but a lot of details remain the same. The term length and payment options remain consistent, and the savings rates are identical (Compute Saving Plans match Convertible RIs, while EC2 Instance Savings Plans match Standard RIs). As it turns out, most of the time you are covering the same usage you would have with RIs and getting the same savings rates back.
Realizing the benefits of Savings Plans
So, when comparing Reserved Instances vs Savings Plans, what potential benefits can the customer realize from this “compression” of usage classifications?
Let’s consider Compute Savings Plans. When you make a purchase, you are effectively committing to an ongoing level of compute usage across a handful of specific AWS services (currently, EC2 instances constitute the vast majority of this usage). EC2 instances qualify for this benefit regardless of their region, operating system or instance family/size. If you purchase RIs instead, you’ll have to make purchases for every family (for non-Linux it’s every family and size!) and operating system combination. It doesn’t take long before this combination list reaches well into the tens or hundreds.
This highlights two clear advantages of Savings Plans. First, they are easier to purchase and manage since you aren’t dealing with a heterogenous set of commitments. Second, you are likely to end up with higher overall coverage, and thus savings, because of the fact that it takes less effort and there is less risk to cover what would have been a long tail of classifications.
Another advantage of Savings Plans is that the core algorithm automatically assigns the discounts every hour to qualifying usage, prioritizing usage with higher savings rates, and thus maximizing ROI.
The key to achieving higher coverage—and thus realizing the most significant benefits of Savings Plans—is, as with RIs, about understanding hourly patterns for the coverable usage and the savings rates for potential commitments.
How Cloudability maximizes your Savings Plans recommendations
When doing this analysis, it’s crucial to consider the commitments you already own and take actions that maximize this investment. This is not necessarily straightforward, since many organizations own a historic set of RIs and newer Savings Plans that will apply to current usage.
Cloudability not only factors in all these commitments as a precursor to determining additional savings opportunities, but also identifies cases where an existing convertible RI can be exchanged (e.g., change the instance family or operating system) to get better utilization and savings. This way, you can increase your total savings without upping your commitment position.
To get a summary of exchanges, navigate to the Reserved Instance Planner and select the EXCHANGE action. It’s worth noting that the Savings Plan purchase recommendations you generate in Cloudability will also factor in the benefit of exchanges.
When generating these recommendations, we believe it’s important to give you the flexibility to tailor a plan to your organization. This starts with letting you choose the exact date window for evaluating usage patterns.
Compare different Savings Plans with Recommendation Support
Cloudability’s recommendations support the two major types of Savings Plans: Compute Savings Plans and EC2 Instance Savings Plans. You will get higher savings rates with EC2 Instance Savings Plans, but higher coverage of usage with Compute (EC2 Instance Savings Plans are just for EC2 and scoped to an Instance Family and Region). Choosing between these two options is an important decision for any FinOps team. Cloudability can make this process a lot easier by clarifying the categories of usage that can be covered and contrasting the savings on offer.
Understand expected impact, then act
After settling on a general approach, the next step is to understand each applicable savings opportunity. Part of this is evaluating the usage that can be covered by additional commitments—and because commitments get consumed or become unused on an hourly basis, this means hourly-level analysis. The detailed view for each opportunity provides a powerful set of insights in one visualization, contrasting your current commitment position and recommended purchase against relevant usage from the selected date window.
By moving your cursor across the histogram, you can instantly identify hourly patterns and validate the “baseline” usage it makes sense to cover versus spikes that should remain on demand. It is worth noting that the “Existing” and “Buys” layers are shown at the discounted price (since this is how you purchase Savings Plans from AWS), while the remaining usage is shown at the on-demand price. In other words, the visualization clarifies what you would pay in total if the recommendation was followed.
Our tip: Many organizations have success purchasing Savings Plans incrementally and reviewing the status on a regular cadence. This can help the FinOps team build confidence that their purchases are having the expected impact.
Finally, before making any purchase, you might find it valuable to assess the cash-flow implications and the estimated savings over the term.
At Apptio, we believe Savings Plans are the future of commitment-based discounting for AWS and expect them to be rolled out to additional services. As there are currently no Savings Plans for RDS or Redshift, we encourage our customers to continue leveraging Reserved Instances to lower hourly rates for this usage. Cloudability’s interactive recommendations are here to help your FinOps team understand how they can be applied to your usage and drive ongoing cost efficiencies.