A Deep Dive Into Cost Savings Offerings from AWS

A Deep Dive Into Cost Savings Offerings from AWS

Sahil Armaan Kumar

In our previous article "An Overview of AWS Instance Purchasing Options", we examined some of the most common payment plan options that AWS provides for purchasing instances. In doing so, we discussed the differences between purchasing On-Demand Instances, Reserved Instances, Savings Plans, and Spot Instances.

Today, we’ll be building upon what we learned in that article and taking a deeper dive into the commitment based cost saving opportunities offered by AWS. Specifically, we will focus on savings plans, reserved instances, and private pricing term sheets. In order to facilitate the comparison of different AWS pricing models, we’ll look at a specific scenario to see exactly how savings would translate across each of the three savings offerings focused on in this post.

Example Scenario

We are a subsidiary of a multinational IT company working on a new project to create a streamlined video compression and distribution software called KwikShare. Our parent company has tasked several teams to collaborate in order to develop this software as soon as possible.

We are currently running 100 dedicated t3 extra large Amazon EC2 instances (a common flexible yet general purpose instance type) to support our dynamic workload requirements. We project the success of this venture and thus anticipate the need to scale up our infrastructure to 300 instances in the next two years, but have decided to begin with purchasing instances at a commitment period of 3 years, leaving us the flexibility to scale up the infrastructure should our prediction come true.

The price of purchasing the required 100 instances On-Demand is approximately $15,184 monthly, or $546,624 for three years, but let’s look at some ways to bring that intimidating number down:

  • Using a Compute Savings Plan for the scenario would cost us $7,533.60 monthly, bringing our total down by over half to $271,209.60 for the three years.
  • An EC2 Instance Savings Plan would instead cost us $6,562.70 monthly, or $236,257.20 for the three years, saving nearly 13% over its compute counterpart.
  • Should we choose to use Convertible Reserved Instances, we would be paying $6,753 per month, totalling to $243,108.
  • Alternatively, with Standard Reserved Instances, our monthly cost would go down to $5,971, reducing our total expense from Convertible RIs by almost 12% at $214,956.

Further, AWS offers a volume based discount to companies who fulfill a certain quota for annual spending on RIs. If our parent company had a yearly expenditure of $500k to $4 million, we would receive a 5% discount on upfront and hourly costs of all future RIs purchased within the specific AWS region while our expenses remained in the discount bracket. This discount doubles to 10% on spending of up to $10 million, and increases further at rates agreed upon by our company and AWS should we spend more than that.

Private Pricing Term Sheet

As a global conglomerate, our parent company spends millions of dollars on AWS every year, thus meeting the $500k years spending quota to qualify for a private pricing term sheet. We also have the resources to employ a highly accurate team of analysts that have determined with full certainty that the company, including all subsidiaries, will spend $7-8 million dollars on AWS in the coming three years.

After some communication with AWS representatives, we negotiated a commitment of $8 million spent over 3 years in exchange for a 12% additional discount on AWS spending across the board. This hefty discount helped cover the additional expense of scaling up the infrastructure of our product once KwikShare took off, as we predicted it would.

Comparing the Purchasing Options

While the flexibility of On-Demand instances can be very tempting, we can see that purchasing instances On-Demand may not be sustainable for all organizations in the long run, especially those with limited budgets. This money could be more efficiently spent for things like product development for example. Even for companies that have larger budgets, growth inevitably increases cloud cost. When cloud spend is large enough a 1-2% difference in margins can mean a lot.

Between RIs and Savings Plans, Savings Plans are often seen as a better choice as they offer savings with more flexibility, and most importantly require less management. However, from the perspective of maximizing savings, RIs are better because they provide signficantly higher savings over On-Demand. The challenge with RIs is balancing flexibility with savings. RIs can be utilized for consistent, long term workloads that are predictable and thus don’t require the extra flexibility provided by a Savings Plan.

With enough resources to spare, investing in budget and workload forecasting for determining an accurate price commitment for a Private Pricing Term Sheet is useful as it provides extra savings on top of those already made through time based commitments including RIs and Savings Plans. However, since expenditure exceeding the commitment price is charged at On-Demand rates, there is a steep penalty for an inaccurate forecast.

So, What’s the Best Option?

Ultimately, each cost saving offering by AWS has its own pros and cons, but the best solution in most cases is a combination of instances purchased under different cost reduction plans. Through such a blended approach, companies can gain the cost reduction from RIs while retaining the flexibility of Savings Plans for dynamic workloads. The attached table shows six real life scenarios where companies using just Compute Savings Plans switched over to the blended approach we’ve been discussing. As you can see, their savings increased by 36-48% across the board, allowing the companies to maximize savings without needing to compromise functionality.

While the benefits of blending different purchasing strategies are clear, selecting the right commitments is still not an easy task. It requires comparing across hundreds of thousands of different combinations of commimtments. That's where tools that can optimize across different commitment types, term lengths, and upfront spends come in handy.

Stay up to date! Subscribe to the Archera newsletter to get updates on cloud offerings and our platform.

Get back to innovating.