#21. Which One Saves More: “Buy two get one free, or buy three get two free”? - ESR is the Answer.

What are Effective Saving rates and why should you care about it?

The way companies measure their commitment savings is broken. Many assume that high commitment coverage and utilization automatically means they're saving money effectively. But is this really true? Let's find out.

How to Determine the Right Amount of Commitment Coverage?

How much are you saving on cloud commitments? Is it 20%? 30%? Or even more?

The question is: how do you measure these savings? Should you look at your RI/SP coverage, or focus on how much of it is utilized?

Is high coverage or utilization the answer?

What if you're running workloads around the clock? In that case, maximum coverage with commitments (RI/SP) would naturally lead to the highest savings. However, this raises questions about the value of cloud computing—if you're using the cloud exactly like a data center, wouldn't it make more sense to consider switching back?

Elasticity Changes The Formula

When you start switching off instances to optimize usage (e.g., downscaling or weekend shut downs of instances), your commitment-based savings naturally decrease. This presents a dilemma: Should you increase usage to improve utilization and justify running resources full-time under commitments? Or should you reduce both utilization and commitment coverage, handling the remaining needs with on-demand pricing? Which approach would be more cost-effective?

Which commitment rate coverage yields the best value?

The answer lies in a simple yet powerful formula: Effective Saving Rates. This metric compares your actual costs under commitments to what you would have paid at on-demand rates.

To illustrate this concept, let's compare two supermarket discounts: "buy 2 get 1 free" versus "buy 3 get 2 free." Using the ESR formula, we can calculate that 1 -(2/3) = 0.33, while 1- (3/5) = 0.40. This shows that "buy 3 get 2 free" is actually the better deal.

Calculating the ESR in Practice

Let's assume you have 20 compute instances running in your environment. Because development instances are shut down on weekends, your usage drops significantly during those times. This elasticity complicates the calculation for coverage. Given a high discount of 65% on RI and an average weekly usage of only 58%, how much should you commit? We'll assume the regular on-demand price is $0.1/hour.

Here is the ESR Calculation for 3 scenarios, 90% coverage (18 RIs Purchased), 75% coverage (15 RIs) and 60% Coverage (12 RI).

Results:

Example A: Despite the higher coverage, you would be better off with on-demand only. The ESR was in negative.
Example B: The coverage is lowered and the utilization is not to the Maximum, yet it yields the highest savings.
Example C: Despite high utilization, the ESR remained lower than Example B because of reduced coverage.

How to Use ESR to Calculate Optimal Commitment / On-Demand Ratio?

That brings us to the final point. While ESR indicates which deals are better, it doesn't directly show the best combination. To find the optimal mix, calculate different combinations of on-demand and commitment options (varying both coverage and utilization), then plot them on a graph or identify the highest ESR value.

While this calculation can be done programmatically using the ESR formula, it's only part of the pain. The complexity comes from having multiple savings combinations—mixing Saving Plans with Reserved Instances, plus choosing between standard and convertible RIs. If you'd rather skip this complexity, consider using ProsperOps, which automatically calculates ESR and structures commitment purchases to maximize your savings rates.

This article was inspired by a talk given Eric Carlin, the founder of ProsperOps

Summary

Neither high coverage nor utilization of cloud rates are reliable indicators of savings. The highest achievable saving rate can be calculated using the Effective Saving Rate (ESR) formula.

ESR is a straightforward formula that compares the savings generated from commitments to their equivalent on-demand rates.

Use this formula to determine which combination of coverage and utilization delivers maximum savings. Since optimizing rates is complex, automation tools are recommended.