Having spoken with hundreds of companies using AWS, I’ve learned that while the nuances of On-Demand and Reserved Instance pricing are better understood than they were a year ago, EC2 capacity planning from a cost management perspective is still challenging for IT and finance teams.
In this post we’ll cover some key points and tips for getting your money’s worth in EC2 without going insane.
Some RI Basics
Before we get started, let’s quickly review some EC2 factoids about On-Demand and Reserved Instance pricing models:
- On-Demand pricing is a “pay-as-you-go” model, typically beneficial for short-term usage, and as such has high hourly usage rates.
- Reserved Instance pricing requires an upfront commitment (one-time fee) with discounted hourly usage rates, and is more of a long-term play, with either 1 or 3 year options available. These reservations also come in 3 flavors – light, medium and heavy – with each flavor offering different discount rates based on how much you pay upfront.
- Heavy reservations require you to pay for hourly usage even if you never spin up the matching instance. For light and medium you only pay the hourly usage for what you actually use.
- No matter what pricing model you choose, you can run all your instances 24/7/365. Having a light reservation does NOT mean you need to use the instance in a limited way.
Now for some advanced EC2 fun!
While going with On-Demand pricing is easy – there is no upfront fee, pay as you go, etc. – it can end up being very, very expensive if used for long term consumption.
On the other hand, while reserved pricing offers amazing discounts, what if you are not planning to use this instance type for a long time. Is it worth it to plunk down that one-time, upfront fee?
Of course, the answer is “it depends”. Are you planning on running this instance for one week or a month? Maybe this will be a 3 month project or perhaps a steady state workload planned for the next year.
No matter what your plans, visualizing all your available options makes it so much simpler.
But which EC2 pricing model is right for you?
The chart below shows the four pricing option for a m1.large instance (Linux) running in us.east-1. The On-Demand blue line, while starting out at $0 cost, quickly rises and after 12 months of 24/7 usage, is the priciest option at $2,074.
On the other hand, all 3 reservation flavors, despite starting with their respective upfront fees, end off the 12 month period at signifantly lower pricng than On-Demand.
The question is, will you be running this instance for the full 12 months? If you are, clearly the heavy reservation offers the most value.
But if you have any doubt about that, you can still easily determine what makes the most sense from a financial perspective.
Let’s take the light reservation (red) as an example. Although you start with a $243 upfront fee, after just 3 months the light reservation breaks-even with On-Demand (blue). This represents the end of your lock-in. And any usage past that point is pure savings.
So if you are confident that you’ll be using this instance for the next 3.5 months, a light reservation is a no-brainer.
Similarly for the medium and heavy reservations. The chart clearly shows the break-even for medium at 5 months and heavy at 7 months, enabling you to easily decide which pricing plan would offer you the most ROI with the least upfront risk.
Forecasting for the enterprise
While visualization is helpful even for small AWS accounts, it becomes critical for businesses with consolidated billing, numerous linked accounts, multiple business units and more. Here, it becomes extremely complicated to determine your actual usage and for which instances RIs make sense. An analysis that factors in aggregated EC2 consumption, along with how RIs are actually applied across linked accounts, while considering your company’s policies for chargeback, is essential.
Fortunately, Cloudyn does all of this heavy-lifting, with clear analytics, accurate recommendations and wealth of other tools and reports, so you can focus on running your business.