Hourly billing on AWS allows you to get your feet wet without overcommitting budgets and other resources in the cloud. As demand increases however, and a fixed number of instances are constantly running, hourly rates can quickly add up, forcing companies to consider volume purchases in the cloud. Reserving a set number of instances one or three years in advance can provide significant savings, up to 60%, compared to on-demand rates.
However, using Reserved Instances versus other payment instruments, such as on-demand pricing, spot pricing and Savings Plans, requires a careful consideration of your business needs in the cloud.
Let’s briefly review all the pricing models to help you understand how each work and how they compare to Reserved Instance Pricing.
This pay-as-you-go AWS pricing model allows organizations to pay for the services that they use, as needed. Compute resources, including AWS EC2 pricing, are assessed by the hour.
AWS offers excess capacity that it has available for any surges in customer demand to offset the loss of idle infrastructure. You can acquire unused spot instances for discounts of up to 90% off the on-demand price.
Customers can commit to spend a specific dollar per hour amount over a one, or three-year term, for discounts up to 72% off on-demand rates. Any usage that exceeds the purchased volume of hours are billed at the regular on-demand rate.
What are the advantages and disadvantages of on-demand, reserved and spot instances?
Your application and workload requirements determine which type of instances will optimize performance. Understanding the wide range of pricing models will help you identify the right mix to maximize savings.
Learn how the fully managed Parquantix solution automatically calculates the lowest cost based on the optimal mix of all pricing models by AWS.
To evaluate which instances and pricing plans are best for your use cases in the cloud, we will take a high level view of your application requirements. We will then dive deeper into Reserved Instances and compare the pros and cons of this pricing instrument.
Choosing the Instance Type for Your Application Requirements
Reserved Instances are most suitable for the following use cases:
- Production environments and applications that require servers to be available 24×7.
- Steady state use cases with predictable workloads.
- Mission critical applications that run on Multi-Available Zone database deployments to ensure higher availability and data stability.
On-demand pricing is more suitable for non-production workloads, such as development, testing, QA, and training. It allows you to pay for the services that you use by the hour, as needed. But if you start running production environments with steadily increasing workloads you will need to consider other pricing instruments.
You can use spot instances if you have urgent computing needs for large amounts of additional cloud capacity. They are ideal for applications that have flexible start and end times that are only feasible at very low computing prices. Examples include test and development workloads with stateless, fault-tolerant, or flexible applications, such as big data, containerized workloads, continuous software integration and deployment, web servers, high-performance computing.
The major drawback to spot pricing is that they can be interrupted at any time with only a 2-minute notification when AWS needs that EC2 capacity back. You need to be able to handle that interruption to take advantage of spot instances.
What factors should you consider when choosing AWS Reserved Instances?
RIs are best for production applications with predictable workloads and high availability requirements. Understanding the different reservation types, attributes, payment options, will help you optimize your cloud deployment with the right RI mix.
EC2 Reserved Instance Attributes
Unused Reserved Instances can be sold in the AWS Marketplace, before their term expires. To be able to sell in the marketplace, companies must have a US bank account.
Identifying the right type of reservation for your needs may become overwhelming given the multitude of choices available. You can start by considering the attributes of RIsto help you decide which type to reserve.
RI attributes :
- Instance type: Instances are available in varying combinations of CPU, memory, storage and networking capacity. For example, m4.2xlarge.
- Platform description: Operating system, such as Linux/UNIX, Microsoft Windows Server, and Microsoft SQL Server
- Tenancy: Default tenancy lets you share the same virtual server without having to interact with others. Dedicated tenancy allows you to run instances on servers specific to your account, for use cases like security and compliance.
- Availability Zone (optional): You purchase an RI to use in a selected Availability Zone. When an Availability Zone is not specified, the RI discount will apply to a running instance of any size (within the same family) in the region.
EC2 Reserved Instances by type and size
Please review our earlier article to identify the right type and size for your needs. Current generation Reserved Instance types, such as r5, c5 and m5, provide higher performance at lower cost than older generation instances. We recommend that you launch new instances with current generation types and upgrade them to newer types when they become available.
RI Offering Classes – Standard vs. Convertible Reserved Instances
You can choose between Standard and Convertible RIs for your application needs:
- Standard RIs allow you to modify Availability Zone, scope, network platform, and instance size (for Linux RIs).
- Convertible RIs can be exchanged for another one of equal or greater value.
The table below compares Standard and Convertible RIs across a range of attributes.
What are the advantages and disadvantages of AWS RI offering classes? (Standard, Convertible and Scheduled)
Convertible RIs provide more flexibility than Standard RIs. Because you can change their attributes and payment options they can help you manage shifts in performance for steady-state usage.
The pros of Convertible RIs can be summarized as below:
- Convertible RIs are ideal for instance types used for edge cases. For example, if you own a g3.2xlarge (Windows) RI and need to move up to g3.8xlarge, the Convertible is an ideal choice.
- They are better for instance types with uncommon sizes and/or for instances running on non-Linux/UNIX platforms. For example, a m4.4xlarge Windows is a good candidate for a convertible RI while a m4.large Linux definitely is not.
- You can convert an RI without penalty.
- The ability to split and join Convertible RIs allows for deeper RI penetration.
You have to consider the following limitations of Convertible RIs:
- Discounts for Convertible RIs are lower than Standard RIs. So, common usage instances, like M4 medium Linux, are not a good fit for Convertible RIs.
- For larger companies or for those with predictable workloads, the pricing for the 1 year convertibles is not very compelling. The 3 year RI is a better option.
- You can use Convertible RIs to take advantage of future price reductions or you can use them to handle usage changes, but you can’t do both.
- Unused Convertible RIs cannot be sold in the Reserved Instance Marketplace.
Other considerations can be based on the OS that you are running:
- Convertible RIs are well suited for Red Hat Enterprise Linux and Windows.
- They are not suitable for SQL Enterprise. The hourly cost is too high, creating too much risk and reducing your savings.
- You can use them as a “hat” for Linux/UNIX, creating a base of partial and no-upfront RIs and then finish with convertibles.
We recommend buying the largest Convertible RI that makes sense for your environment. When your needs change, you can split it into smaller sizes. This prevents you from incurring costs if you need to upgrade your resources.
Convertible RIs require advanced knowledge, you need to make sure that you fully understand the complexities before purchasing them.
Scheduled RIs are available to launch within the time windows you reserve. This option allows you to match your capacity reservation to a predictable recurring schedule that only requires a fraction of a day, a week, or a month. According to AWS, Scheduled RIs provide a 5-10% savings over On-Demand rates.
Scheduled RIs have a limited use case, for predictably scheduled operations that will run for at least one year. Many companies will look for more than 10% savings for a year-long commitment.
Scheduled RIs are also not flexible. They are only available in a few regions and for very few instance types.
Parquantix determines the best RI type based on each customers’ application needs, usage patterns and workload volume. Contact us to schedule a 30-minute consultation to find out how we can help you.
RI Payment Options
Another factor to consider in choosing RIs is how much upfront payment you can afford pay AWS to take advantage of the deep discounts.
You can choose from these payment options:
What are the advantages and disadvantages of all upfront, partial upfront, and no upfront payment options?
You have to compare the potential savings over the RI term, against the cost of the upfront investment. In our experience partial upfront provides better value.
Considering the pros and cons of each payment option will inform your decisions.
Pros & Cons of No Upfront Reservations
- No upfront option allows you to preserve cash for other critical capital and operating expenditures
- No upfront reservations are more appropriate for t2 general purpose reservation types, cost difference is small compared to other options.
- They are extremely difficult to sell in the AWS Marketplace. Because the upfront price is $0, you cannot discount it to entice a sale. If AWS lowers their prices, as they frequently do, you will get stuck with the.
Pros & Cons of Partial Upfront Reservations
- Partial upfront reservations have a very low hourly cost
- Unused RIs can be easily sold in the AWS Marketplace
- You will need to be prepared to make an upfront cash investment
Pros & Cons of All Upfront Reservations
- There is no hourly charge for all upfront reservations
- They provide the greatest discount compared to on-demand rates
- Unused RIs can be easily sold in the AWS Marketplace; and.
- However, compared to a partial upfront reservation, this option provides very little additional cost savings. You have to weigh the benefit of the discount against the cost of the upfront cash investment. This criteria make all upfront RIs a better investment for enterprises with large workloads.
Reserved Instances for Amazon RDS (Relational Database Service)
Reserving dbase services on AWS provide significant savings. RDS RIs even provide better savings than EC2 instances. The savings are especially high for Amazon Aurora or MySQL.
But what happens when you purchase an RDS reservation and then six months later, you need a bigger database service? Therein lies the main issue with RDS reservations. Although they provide greater cost savings than EC2 reservations, you must ensure that you are able to use an RDS RI for at least one year, since scalability requirements for database services may increase substantially.
Another limitation of RDS RIs is that there is no marketplace for selling RDS reservations. So, if you don’t use the RDS RI before its term expires, you will get stuck, counteracting most of the cost savings that you initially planned on.
Reservations can reduce costs significantly, but they must always be matched against actual usage. You must diligently monitor your reservations against usage or your savings can be easily wiped out. The added complexity of identifying the right instance types with the right attributes and payment options can quickly become overwhelming. Fortunately, Parquantix can do all that for you. Our automated optimization tool manages the complexity and our data-driven algorithms decide which risks are worth taking.