Layering: A Mille-Feuille Strategy for Optimal Cloud Commitments

Managing cloud costs is somehow like investing in the stock market. Imagine putting your entire budget into a single stock on a single day, betting on a fixed decision that locks you in for the long term. It’s risky, rigid, and rarely optimal.

Instead, smart investors use strategies like dollar-cost averaging: spreading investments over time to mitigate risk and adapt to changes. While the analogy isn’t perfect—cloud spending comes with its own unique dynamics—the core principle of balancing commitment with flexibility remains the same.

This article explores how the concept of Layering can help organizations stagger cloud commitments, reduce financial exposure, and adjust to evolving business needs.

Whether through a straightforward manual approach or leveraging more advanced data-driven techniques, layering offers a practical path to better risk management and eventually higher net saving rates.

The Limits With One-Off Commitments

Major cloud providers, including AWS, GCP and Azure all offer discounts on compute usage in exchange for long term commitments (e.g. Reserved Instances and Savings Plans at AWS, Committed Use Discounts at GCP, Reservations and Savings Plans for Compute at Azure).

These agreements typically span one or three years and reward organizations for committing to a specified level of usage. By opting for a large, one-time commitment, businesses can immediately unlock the full discount associated with their commitment level, maximizing short-term savings.

While this straightforward approach can work, especially for companies with stable and predictable workloads, it comes with its limits and challenges.

1. Evolving business needs: businesses and the cloud infrastructure that support them are rarely static. Changes in infrastructure strategy, the launch or sunset of product lines, expansion to new market segments and geographies can all disrupt initial forecasts.
2. Seasonality: resource usage often fluctuates due to seasonal trends - retail holiday peaks, travel vacation period surges, or recurring workloads like monthly billing processes. Fixed, one-off commitments don’t accommodate for these variations, leading to overpayment during quiet periods and high on demand prices during spikes.
3. Tight renewal windows: when a large commitment expires, organizations must renew immediately to avoid reverting to costly on-demand pricing. This tight window leaves little time for adjustments, often rushing businesses to get locked into new commitments based on outdated forecasts.

The Layering Approach

Layering is a strategic approach to cloud commitments that promotes flexibility. Instead of making a single, large commitment in one go, an organization can distribute their commitments over time in smaller, incremental layers, each layer with its own start date and term length.

This staggered approach offers key advantages:

1. It reduces the risk of over or under-committing and provides opportunities to adjust commitments in response to changing business needs.
2. As each layered commitment expires, an organization has the option to renew it, resize it or let it expire, maintaining the ability to align their commitments with their actual usage patterns and evolving infrastructure.
3. If managed properly, layering allows to align commitments levels with seasonal trends even where seasonal periods are shorter than the minimum commitment term.

How does layering look like - a concrete example

Consider an organization forecasting a need for 1,000 compute units per year. Instead of committing the entire capacity upfront for a one year term, they could break this into four layers:

• Layer 1: Commit 25% (250 units) for a one-year term starting in January.
• Layer 2: Commit 25% (250 units) for a one-year term starting in April.
• Layer 3: Commit 25% (250 units) for a one-year term starting in July.
• Layer 4: Commit 25% (250 units) for a one-year term starting in October.

With this structure, the organization avoids having all commitments expire simultaneously. Instead, they renew or adjust one layer every 3 months, allowing them to account for changes in usage needs.

Staggered Expiration Profiles

The key benefit of layering is the flexibility it creates through staggered expiration profiles. Rather than facing the pressure of renewing an entire commitment block when it expires, organizations can spread out renewal decisions across multiple layers. This incremental approach reduces financial and operational risk by providing regular opportunities to reassess needs and make adjustments.

A Gradual Build to Savings

Layering introduces a trade-off: the full extent of savings isn’t realized immediately, as commitments are spread out over time. This means that during the ramp-up phase, only a portion of the workload benefits from discounted rates. However, this gradual approach allows businesses to stay flexible and better align their commitments with real-world usage in the long term.

Implementing Layering in Practice

There are many ways to implement layering, ranging from simple, manual methods with predefined schedules to more advanced, data-driven approaches. Each method comes with its own trade-offs between complexity, flexibility, and efficiency.

Manual Approach

The manual approach is a straightforward way to implement layering, making it accessible to organizations of all sizes. The idea is to break down the total commitment target into smaller increments and spread them over time. A key decision in this approach is determining the frequency at which layers are committed, as it directly impacts the balance between flexibility, ramp-up speed, and operational effort.

Higher frequency layering (e.g., committing every month) offers greater flexibility and reduces the risk of over-committing, but requires more frequent forecasting exercises adding operational complexity.

Lower frequency layering (e.g., committing every 3 months) accelerates ramp-up to full savings and reduces operational overhead but provides less flexibility to adjust commitments as usage patterns evolve.

Organizations adopting this method must find the right balance between ramp-up speed and flexibility, depending on their workload predictability and resource planning capabilities.

Data-driven Approach

For organizations seeking a more sophisticated strategy, machine learning and in particular reinforcement learning can greatly enhance the effectiveness of layering. Reinforcement learning models are designed to continuously learn from new data and come up with dynamic strategies that optimize for a specific outcome.

By simulating various commitment scenarios, these models can determine the optimal size and timing of each layer to minimize costs while accounting for uncertainties.

The Opsima optimization engine

At Opsima, we specialize in helping organizations optimize their cloud costs with advanced algorithmic tools. Our optimization engine leverages reinforcement learning to implement dynamic layering, alongside other optimization levers, to deliver maximum savings while minimizing risk.

By adapting to your unique usage patterns and business needs, our solution ensures you stay flexible and efficient in an ever-changing cloud environment. If you’re ready to take your cloud cost management to the next level, reach out to learn more about how Opsima can support your goals.

‍