Cloudability logo
Listen to our AWS re:Invent 2018 Reaction
Optimization

Implementing a Strategy for Exchanging Convertible Reserved Instances

By James Jackson on November 14, 2018
In addition to providing functionality to optimize the purchasing, optimizing and ongoing management of EC2 Convertible Reserved Instances (RIs), Cloudability now generates Convertible RI Exchange recommendations to further optimize your cloud infrastructure portfolio.

All successful organizations are constantly undergoing change. Convertible RIs, by nature, cater to that fact. Convertible RIs offer a path to achieving savings over on-demand resource usage with less risk than a traditional reservation — if properly managed. While Convertible RIs do not offer quite the same discounts as a standard RI, the enhanced flexibility provides an opportunity to be more aggressive for even greater savings.

Whether used independently or in conjunction with standard RIs, the flexibility and financial benefits of Convertible RIs is clear. However, developing and implementing an appropriate strategy is often less straightforward. Like most things within cloud infrastructure, IT and technology in general, an appropriate strategy depends on your use case and where you are in your cloud infrastructure journey.

First introduced in the Fall of 2016, convertible RIs offered the potential to achieve near the savings of standard RIs (~45% vs ~40-60%). Their flexibility, and therefore usability, was greatly enhanced in Q4 of 2017 when users were granted the ability to merge and split convertible reservations. Unfortunately the flip side of the expanded flexibility was a jump in the number of permutations of exchange possibilities, making it that much harder to identify the most optimal choices.

Like the rest of the Cloudability solution, EC2 Convertible RI recommendations remove the guesswork, providing clear, easy-to-execute recommendations with expected outcomes, making large or even frequent exchanges manageable.

Understanding Savings and Utilization Thresholds

Before looking at a few strategies, let’s review the concepts of Savings Threshold and Utilization Threshold as they apply within the Cloudability Reserved Instance Planner.

Savings Threshold: Savings threshold limits the recommendations you are presented with to those that are at or above a lifetime savings rate that you set. For instance, if a savings threshold of 35% is selected, any recommendations will save at least 35% over on-demand resources between now and the end of the RI term.

Utilization Threshold: Utilization threshold limits recommendations to those that are projected to meet or exceed a certain level of utilization. For example, if a user sets a utilization threshold to 90%, only reservations expected to have utilization at-or-above 90% are presented recommended targets.

Think Ergonomically — Fit the Strategy to Your Organization

When it comes to Convertible RIs, there are offensive and defensive strategies, and good reasons for adopting either. For instance, if you are in a hyper-growth, born-in-the-cloud organization or you work on the customer-facing SaaS infrastructure of your organization, a more aggressive offensive strategy is more in line with helping you achieve success. Conversely, if you have completed migration of your stable, back-end enterprise workloads from your datacenter to cloud infrastructure, a defensive strategy is likely more warranted.

Doing Your Thing, Only Better — Steady State Workload Strategy

One common use of Convertible RIs is to migrate well-characterized workloads (at any given time throughout their lifecycle) to the most favorably-priced instance families. In this type of scenario, one effective strategy includes optimizing with a high utilization threshold (e.g. >90%) but with a lower savings threshold (e.g. 0-20%). This type of strategy:

Embracing Change — Hyper Growth Workload Strategy

Another scenario when Convertible RIs are regularly utilized involves coverage for workloads with exceptionally high growth rates over a protracted time frame (e.g. >20% monthly growth for >1.5-years). There are many examples of born-in-the-cloud SaaS offerings that exhibit this behavior, including gaming, social/community platforms, auction, streaming media, finance and education, along with many others. While you still want to take into account negotiated pricing for specific platforms and the cost advantage of reserved instances, successful convertible reservation strategies for these scenarios centers around the ability to take advantage of the latest technology and accommodate quickly changing requirements.

Because of rapid traffic, and therefore cloud resource growth, organizations or departments within organizations responsible for these workloads are less adverse to resources maintaining a little extra ‘head-space’ and regular changes. For these reasons, compared to steady state type workloads, it is better to generate recommendations with a lower utilization threshold (e.g. 60%) and a higher savings threshold (e.g. 45%). This strategy works well for these use cases because:

This type of strategy typically involves more frequent exchanges and upfront can lead to issues with reconciliation of cost through multiple layers of parent/child instance inheritance.

New to Convertibles Strategy? Think Chess, Not Checkers

If your organization is just getting started with Convertible RIs, remember it’s not a one-and-done process. Changes and the resulting improvements will take goals, strategy, planning and iteration. As the old saying goes, “How do you eat an elephant? One ‘byte’ at a time.”

I have found that although virtually everyone is in favor of progress, the individual changes that lead to progress are sometimes harder to get on board with. Take it in small achievable chunks, allowing your organization to get comfortable by celebrating some wins. This way will give you more opportunities and time to iterate, adjusting strategy with less waste.

This is pretty much explicit in the name Convertible RIs — the idea is to convert them as things change. The likelihood that your organization will remain the same over the next 1-3 years is slim and therefore it is also unlikely your infrastructure needs will either. Whether your goal is to make sure you can stay on the latest technology, make sure you have resources to expand economically without buying more than you need or your organization is just “learning” your way into the optimal infrastructure type, the flexibility of convertibles enables you to more effectively manage your risk.

When getting started with convertibles one reasonable strategy to tune for generating exchange recommendations includes no-upfront reservations with moderate savings (e.g. 30%) and higher utilization thresholds (e.g. 90%). Doing so will help minimize the number of actions and shortens the break even point over absolutely maximizing savings while you are able to better characterize the needs and behavior of workloads in the cloud, learning your way into longer term convertible reservation strategies.

To adjust over the short term:

No matter your workload or where you are in your journey with convertible reservations, cloud resource and cost optimization is a continual process. Your needs, technology and prices are all dynamic by nature and therefore your optimization plan must be. With Cloudability you have a team of cloud cost optimization experts and the industry’s leading suite of intuitive optimization tools, including Convertible Reserved Instance Exchange Recommendations to help you achieve your goals.

Do you have a different use case where Convertible RIs have been particularly useful? Or questions about the next step to take? We’d love to hear from you. Contact us.

——

If you’re not currently a Cloudability user and would like to learn more about Convertible RIs, sign up for your free trial.

Being in the know feels great