Enterprise companies can save up to 70% on compute costs by committing to short- or long-term usage. So why aren’t all businesses using Reserved Instances for their applications? The reason is that purchasing, renewing, and managing the usage or coverage of Reserved Instances is extremely complex.
What are Reserved Instances?
Reserved Instances, or RIs, are a financial commitment or billing discount that a company can get when they adhere to a specific level of instance usage. Organizations can secure significant discounts when compared to standard “on-demand” cloud computing prices. In most cases, the terms of these deals are between one to three years for a low rate. Business applications with steady and predictable workloads are ideal for RIs because they allow for better coverage.
What are the challenges with managing and purchasing RIs?
Managing and purchasing RIs is a real challenge. So much so that some organizations have entire teams or roles dedicated to the task. The six most important factors you need to consider when it comes to investing in RIs are:
- There are hundreds of instance type options
- Certain RIs are only available in certain regions or zones
- Visibility across multiple accounts is not natively provided by the cloud provider
- Terms or Length of Commitment can range from 1-3 years
- Payment Options include no upfront, partial upfront, or all upfront costs
- Offering Class (AWS & Azure only) will determine the flexibility of applying your RIs
Overwhelming Instance Type or Attribute Options
Each cloud provider offers hundreds of unique instance or virtual machine types that range from general purpose to memory intense types. Instance Attributes include instance type, region, tenancy (type of hardware, single or shared), and platform (OS type). It can very quickly be overwhelming trying to determine what are the right RIs for your cloud workload.
Availability of RIs within Regions or Zones
When it comes to availability, I’m referring to the availability of Reserved Instances within a specific Availability Zone (AZ) and/or Region. Not all instance types are available within every Region or AZ. It’s important to understand instance type availability when designing your application and is yet another reason RI management is too complex to manage manually.
Visibility Across Multiple Accounts
Cloud providers do not natively provide complete visibility across your entire cloud organization, accounts, and regions for Reserved Instances. Individual accounts can purchase Reserved Instances, but they are not visible to other accounts for utilization. You need a separate solution to get holistic visibility across multiple accounts.
When purchasing RIs you need to choose the terms or length of commitment. Again, this can range between and three years. Sounds simple enough, but it requires that you’re confident about the consumption needs of your cloud workloads and where they will run for that range of time. Are you?
For AWS and Azure, the available payment options range from, no upfront cost; all upfront, meaning paying for your RIs in full to start; or partial upfront, where you spend a little to lower your monthly cost. GCP only has one payment option: no upfront.
Offering Class is broken down into two types: Standard and Convertible/Size Flexibility. Standard provides the highest savings but cannot be exchanged. It is bound to one instance family on the same operating system. Convertible Instances provide a lower discount but can be exchanged or modified.
Different cloud providers
The top three cloud providers offer a variation of Reserved Instances or committed use discounts. For all providers, it’s important to note the “no cancellation policy.” Once Reserved Instances are purchased, they are yours, whether you use them or not.
Azure offers up to a 72% reduction in your cost when utilizing Reserved Instances.
When purchasing a Reserved Instance, Azure has a feature turned on by default called “instance size flexibility.” Instance size flexibility will apply the reservation discount to other VMs in the same size group but only within a specific region. The other option is called Capacity Priority, this assures the prioritization of data center capacity deployments and your ability to launch a VM when you need them. The tradeoff is that it is only available when the reservation scope is a single prescription.
AWS offers a few different was to save on reservations, from Reserved Instances to Savings Plans. Purchasing standard Reserved Instances will provide you up to 72% cost savings and follow the same rules outlined under Offering Class (Standard) (written above). Limiting you from exchanging them or moving to another OS type.
The other option is purchasing convertible Reserved Instances that AWS offers, which typically give you just a 45% discount, but more flexibility. Convertible RIs allow you to change the instance family associated with a Reserved Instance any time, as long as the RI is of equal or higher value than the original converted RI.
AWS has created an open market called the Reserved Instances Marketplace to sell Reserved Instances that are no longer being used to their recommended capacity. It’s also a place to purchase Reserved Instances at a steep discount. Other cloud providers do not have a secondary market to elevate unwanted Reserved Instances.
AWS Offers Savings Plans, Should I Buy Them Instead of RIs?
We won’t be going in-depth into Savings Plans but there are important distinctions that must be made between the two. AWS released three types of Savings Plans in 2019 to try and make things easier to manage for their customers, each plan was created with a different customers’ need in mind. The three types of Savings plans are, compute Savings Plans, EC2 Instance Savings Plans, and then Amazon Sagemaker Savings Plans.
Check out this 2019 blog for a good overview of the Savings Plans, but note that AWS has since updated and added new features since this article was published including the addition of SageMaker plans. AWS Compute and EC2 Savings plan features haven’t changed.
The main difference between Savings Plans and Reserved Instances is that Savings Plans are an hourly commitment of a specific spend amount with up to 66% off on-demand pricing; while Reserved Instances are a billing discount applied when you adhere to a specific level of instance usage for 1-3 years. For Savings Plans, any compute usage that is up to that commitment will be at the discounted rate, but if the usage goes over then it falls back to the on-demand price.
GCP offers a Committed Use Discount on Compute Engine usage. Cost savings can range from 57% for most machine types, including GPUs or up to 70% for memory-optimized machine types. Committed Use Discounts are flexible and cover a wide range of resources that can be applied to a single project or multiple contracts/projects by enabling the shared discount option when purchasing committed use discounts. The only payment option available for committed use discounts is no upfront costs.
GCP does break down committed use discounts into two types of commitments: spend-based commitments or resource-based commitments, similar to what AWS offers with Savings plans.
Spend-based commitments provide a discount in exchange for your commitment to spend a minimum amount and are only available for the following services: Cloud SQL Google Cloud VMware Engine, Cloud Run, and Google Kubernetes Engine.
Resource-based commitments are discounts in exchange for your commitment to use a minimum level of Compute Engine resources in a particular region.
How Turbonomic Can Help
Turbonomic takes a performance-first approach with its continuous cloud optimization, which is why our customers trust the platform to execute and automate our resourcing actions ensuring they only consume what they really need. In addition to performance assurance, our customers want to take advantage of the additional cost savings that cloud providers are offering through RIs and Committed Use Discounts. As we’ve discussed above, these are extremely complex to manage, which is why Turbonomic continuously and automatically considers RIs as part of its compute scaling analysis: from instance types and the availability of RIs to terms and payment options. The software prevents resource constraints while also ensuring that Reserved Instances are being used to their fullest capacity.
With Turbonomic you’re not only assuring application performance, but you’re making the best use of RIs—from your current RI coverage to your RI utilization. Available today for AWS and Azure (with GCP coming soon), Turbonomic’s automated RI-aware scaling and purchase recommendations helps in the following ways:
- Helps you purchase the first order of reservations
- Optimizes the use of your current reservations
- Helps you re-purchase soon-to-be-expired reservations
- Continues reservation purchases for growing environments
Check out Try Turbonomic where you can gain access to a LIVE Sandbox environment allowing you to learn all about the benefits of our platform’s continuous optimization capabilities.