Amazon has announced Savings Plans – a pricing and discount plan for EC2 and Fargate. We’ve often thought the Reserved Instance (RI) model to be overly complex. In many cases, it is more disadvantageous than it is useful. That’s why we’ve never created automation for RIs.
For an organization that preaches the power of on-demand, our main objection to RIs is the lack of flexibility. RIs requires you to commit to a specific instance type for 1 to 3 years. One of the most fundamental promises of the cloud is that it is truly on-demand and elastic. You can deploy any cloud services at your fingertips and the freedom to turn them on and off as you see fit and to pay accordingly.
Amazon has adapted over the years to unleash people from this burden by permitting the resale of unwanted RIs on a secondary marketplace. Also, Amazon offers the ability for organizations to buy convertible RIs that permit their buyers to change the Availability Zone, instance size, and networking type. Both of these options bring with them increased discounts but also a great deal of complexity in the purchasing and ongoing management of the Reserved Instances.
From there, it has sprouted several tools and consultants whose sole raison d’être is to untangle the spaghetti of billing and compute confusion unleashed by these programs, in many cases undermining the cost efficiencies that were intended by Reserved Instances in the first place.
More pertinently, organizations who figure out how to use RIs effectively see a few of the benefits of adjusting capacity to meet demand as they’re mostly locked into paying for continuous usage. There are ways around this with striking the right mix of Reserved Instance, On-Demand (and often Spot) but they require complicated calculations and implementations.
At GorillaStack, we’ve always felt that EDP (Enterprise Discount Programs), whereby an organization can receive a discount for an overall financial incumbency is a far better way to reward an organization’s ongoing commitment to AWS without burdening the end user with onerous and expensive complexity. EDPs are usually reserved for organizations spending hundreds of thousands of dollars a month on AWS and that’s why we welcome this “EDP lite” or “EDP hybrid”.
AWS Savings Plans allow organizations to take some (but not all) of the simplicity of on-demand cloud while leveraging the benefits traditionally reserved for RIs and EDPs.
With Amazon’s announcement of their new pricing program, a lot of organizations are asking how to distinguish between Reserved Instances and Savings Plans.
So what exactly are they, should organizations change their approach to Amazon Web Services cost management? What’s the best way to think about Savings Plans in the mix of Reserved Instances and other cost optimization strategies including elasticity automation, rightsizing, spot, and EDPs?
There are 2 types of Savings Plans: Compute Savings Plans and EC2 Instance Savings Plans. Compute Savings Plans can be applied automatically regardless of family, size, region, etc whereas EC2 Savings Plans require you to commit to the same family in the same region. Both require commitments of either 1 year or 3 years.
Compute Savings Plans can be used for EC2 instances and Fargate. The program seems to be designed as a hybrid between AWS’s Enterprise Discount Program and AWS’s longstanding Reserved Instances pricing structure.
Essentially you can commit to a predetermined amount of ongoing compute for either 1 year or 3 years. In return, you receive a discount for that commitment. In the case of Compute Savings Plans, unlike Reserved Instances you don’t have to apply that commitment to a certain type of instance or family of instances, you’re simply agreeing to spend that much on overall EC2 usage or Fargate usage.
To be clear, it doesn’t matter which type of instances within EC2 or Fargate that you spend the commitment on, nor does it matter how many hours you run the services for.
In the case of EC2 Savings Plans, they’re a lot more similar to the classic Reserved Instance, requiring an organization to commit to an individual instance family in a specific region. They differ from RIs in as much as you can change the Instance Type within the family.
The discounts for AWS’s Savings Plans program vary depending on the length of your commitment and the plan you’re committing to. Like RIs, Amazon offers 1 and 3 years terms with No Upfront (pay monthly), Partial Upfront (pay some at the beginning and some monthly), and All Upfront (pay for everything right away). The discounts scale with your commitment but can be up to 72%.
At the time of launch in November 2019, you can expect the following discounts:
|Unit||Reserved Instance||EC2 Instance Savings Plan||Compute Savings Plan|
|Average 1 Year Discount||38%||29%||29%|
|Average 3 Year Discount||58%||58%||51%|
|Instance Size||Fixed (except Linux)||Flexible||Flexible|
|Geography||1 Region||1 Region||Flexible|
|Service||EC2 / RDS||EC2||EC2 / Fargate|
You can learn more about reserved instances in this blog post before you commit.
We believe so, but even with Savings Plans, it’s important that you plan correctly so that you make the correct commitment. If you’re coming off of RIs (or even if you’re not), make sure that your organization takes full account of what an instance scheduler for EC2 can provide when you calculate savings through elasticity and maybe consider using a tool like GorillaStack’s Cost Optimization to drive those savings.
Don’t get caught out by investing in 24/7/365 prices or specific instance types when you are now much freer to take advantage of AWS’s on-demand model. Bear in mind that you don’t have to buy your whole year upfront with on-demand. With Savings Plans, you may want to consider drip-feeding your prepurchases on a month-by-month basis as you become more comfortable with this new model of thinking about AWS discounting. If you’re looking for an alternative, you can also consider spot pricing.