How to Balance Flexibility & Savings in Your AWS Reservation Management Strategy

Getting the most out of every dollar you spend on cloud matters. FinOps teams are under pressure to deliver more with less. As pressure mounts, organizations have a critical opportunity to understand, track, and optimize their cloud spend and commitments for flexibility alongside savings to deliver the best possible results for their teams and shareholders.

The problem

The cloud has always offered incredible flexibility and elasticity for organizations to only consume what they need. Reservations, at the most basic level, have been a way for organizations to trade some of that flexibility for increased savings. While the exact details have evolved (starting in 2009 with Reserved Instances for EC2 allowing organizations to commit to a specific resource type for a deep discount, then Convertible Reserved Instances in 2013 offering additional flexibility with a slightly lower discount, and finally Savings Plans in 2019 with a commitment to hourly spend levels instead of resources), the fundamental trade-off between future flexibility and current savings has been at the core of reservation management.

There’s always been a tension between the savings you can get from commitment-based discounts and the flexibility organizations want to maintain for the future. Historically, organizations have thought of flexibility as a conceptual consideration instead of a hard metric to track and optimize for. It’s often just below the surface for most organizations; they think about it without realizing it and without measuring it. Flexibility most often manifests in the form of some percentage of compute usage remaining On-Demand. The fear of overcommitment has led companies to leave some money on the table in the form of On-Demand spend.

Defining flexibility

Flexibility is all about preserving your optionality for what may come in the future. Before we go any further, lets actually define flexibility. Here at Apptio, we define it as how much you can reduce your compute spend without your commitment dollars going to waste. While the formula we use is quite complex (we consider every commitment across your organization and your associated usage patterns), you can think of it as essentially the following:

Flexibility = 1 – (Minimum Commitment Level)/(Compute Usage)

You should think of the numerator (“minimum commitment level”) as the On-Demand equivalent of the minimum-level commitment that you’d have if you actively manage and reduce your current commitment as much as possible. Different reservation types have different mechanisms built in, allowing you to control those commitments. Taking advantage of those options (e.g., refinancing your current commitments to reduce your hourly commitment by extending the duration) along with constructing your portfolio in a way which maximizes flexibility ensures you can minimize your compute commitment as much as possible. By employing these strategies, organizations can dramatically reduce their hour-to-hour commitment. When the need for compute drops, their hourly commitment can shrink alongside it.

The denominator is simply your current total compute usage as a reference point to normalize your usage.

Flexibility and savings: a trade-off no more

For most organizations (especially those with a heavy AWS Savings Plan investment), flexibility means how much of your spend is On Demand/Spot, which puts savings and flexibility at odds. Most organizations think of this as, “For every extra dollar I commit, I lose some flexibility because I have made a long-term commitment to spend at this level on AWS.”

Instead of thinking of savings and flexibility as a direct trade-off, it’s better to think about how they can complement each other. With a fresh approach to commitment-based discounts, you can construct your portfolio to get both results. With a flexible portfolio, you can use three-year instruments to achieve a maximum discount rate because you have the flexibility to adjust your coverage levels to meet the compute demand of your organization. Building flexibility into your reservation strategy reduces your risk of overcommitting, so with that flexibility, you can be comfortable pushing that coverage higher and delivering even greater savings for your organization.

How you can achieve both savings and flexibility

Hopefully by this point, you’re convinced that an approach to reservation management that maximizes savings and flexibility together is the optimal approach, and the only remaining question is “how?” There are a variety of approaches organizations take to incorporate flexibility and optionality, but not all approaches provide the same results.

A passive approach

Some organizations take a passive approach and are OK with leaving some portion of their spend On-Demand. They are willing to overpay for some of their usage in exchange for the flexibility that comes with it and the ease of buying a Compute Savings Plan, and they call it a day. This approach tackles the low-hanging fruit. It’s a low-effort approach, but it also leaves money on the table.

Leverage Spot Instances

Other organizations take a similar approach as above to commitment management but instead focus on paying less for the consumption above and beyond those commitments by leveraging Spot Instances. Spot Instances are excess AWS capacity offered at a discount relative to On-Demand rates. However, AWS can reclaim those instances when needed with only a few minutes notice.

To effectively run applications and leverage Spot Instances, application teams must go through a re-architecture process to ensure their applications are fault-tolerant. It’s a potential path to maintain flexibility, but it yields unpredictable results. Spot market pricing is driven by excess AWS capacity and demand for Spot Instances, which means it can fluctuate and may not pay off as you hoped. With such dependency on the Spot market pricing, many organizations are recognizing better approaches that can deliver consistent results and don’t require their engineering teams to stop innovation work to re-architect applications to leverage Spot Instances.

If you look back through Spot Instance pricing, you’ll quickly see how Spot pricing can change and increase. For example, in April 2023, organizations saw Spot Instance pricing reach the same rates as On-Demand usage, meaning these organizations missed out on savings during this stretch. In many regions, Spot instances are still more expensive than reservation rates and the organizations spent engineering bandwidth to enable Spot-readiness without being able to deliver the cost savings they were hoping for.

Transact on the Reserved Instance Marketplace

AWS provides a marketplace to buy and sell standard Reserved Instances. In theory, when demand for compute goes up, you can buy more Reserved Instances from this secondary marketplace to cover that usage at potentially a lower price or shorter term than the original commitment. When your compute usage drops off, you can sell those Reserved Instances on the secondary market to avoid them going to waste. Individual organizations can leverage the Marketplace, or you can turn to third-party tools to transact on your behalf. Sounds pretty good, right? Well, it’s a little better in theory than in practice.

If you are counting on the RI Marketplace as your flexibility strategy, you are taking an external dependency on the actions of others in the market, inherently betting on the liquidity of the RI Marketplace in the future. You are counting on a liquid market for the specific reservations you need (e.g., a specific resource type, region, operating system, etc.). You also need it to be liquid at the right time — meaning you need available supply of the types of reservations your organization uses when your compute demand is increasing and sufficient demand for those reservation types when your internal demand is dropping. Think you’re the only one with seasonal spikes in Q4? Think again …

Even when you rely on third parties to transact on your behalf, they’re still limited by the same market realities. Those tools need to take that into account to ensure they don’t get you overcommitted, leading you to take them up on those buy-back guarantees. The result? Lower coverage and lower savings because they have to hedge their exposure.

Portfolio approach with Convertible Reserved Instances and Savings Plans

If the approaches outlined above (and the associated risks) don’t sound like how you want to build in flexibility for your business, we recommend a portfolio approach leveraging the benefits of multiple types of reservations, both Convertible Reserved Instances and Savings Plans. Leveraging the benefits and available management strategies of each type of reservation allows you to achieve savings and flexibility together. This approach isn’t without its challenges either; it requires intentional portfolio construction and active management of these reservations to achieve the savings potential. This active management approach can deliver incredible results. However, it can only work effectively when paired with automation to ensure that the balancing and matching of those reservations to cover the resources running throughout your estate.

Work with the experts

If you’re ready to start pursuing increased flexibility and savings, reach out to the Apptio team and let Cloudability Savings Automation manage your AWS compute reservations and build out the optimal reservation portfolio for you. Cloudability Savings Automation doesn’t just automate the reservation purchasing you do today, it flips that approach on its head in order to deliver results you couldn’t accomplish on your own. Cloudability Savings Automation constructs a reservation portfolio by leveraging multiple kinds of savings instruments to deliver the best of all worlds — without taking any external dependencies (e.g., the Reserved Instance Marketplace).

Cloudability Savings Automation’s approach of active management, portfolio construction, and built-in flexibility allows us to deliver optimal outcomes for your organization, no matter what future challenges or changes may come your way. We don’t just purchase commitments for you and hope that still fits your needs in a few years; we actively manage to dramatically increase savings, reduce risk of overcommitment, and deliver incredible results. Our approach allows us to reshape the commitment curve by scaling commitment up and down to match your demand. Savings instruments no longer have to follow a “pick a commitment level; use it or lose it” approach. With Cloudability Savings Automation, your perspective on savings instruments will quickly become “manage it or lose it” — and there’s no one better to manage it for you than Apptio.

Want to learn more? Check out this whitepaper to learn more about other cost-saving strategies.