Pricing and cost structure of AWS
With the vast variety of services provided by Amazon Web Services, the prediction and management of cost for large deployments can be very complex. AWS has a pricing structure based on 3 basic points. These are:
- Pay as you go
- Payless by using more
- Save when you reserve
I will explain each of these points for your better understanding.
Pay As You Go
AWS leases computing resources to turn capital expenses in operating expenses and it lets its users pay for them hourly.
You should leverage “Pay as you go” only for the workloads that have unexpected peaks or scalability needs. This is because AWS has an On-Demand pricing scheme which is expensive even for small workloads.
Pay Less By Using More
You are discounted on your total cost when you use specific service features and also when you spend more than $500,000 on AWS upfront.
You establish yourself in the Amazon ecosystem when you increase your usage of AWS. You must balance your desire for discounts with the need to maintain a healthy, multi-cloud strategy.
Save When You Reserve
Amazon EC2 is the basis for most of AWS, and it provides discounts of around 30% to 50% if you reserve instances in advance. (approximately 1-3 years in advance)
You must consider the number of workload on-premises before purchasing reserved instances. It will be cheaper if you use the same effect unless there is a need for any flexibility provided by the cloud.
Ways in which AWS shows your cost structure
AWS tailors its cost in a well-structured manner and provides several cost datasets for a better understanding of your AWS usage. AWS Billing and Cost Management provide the following datasets for better cost management of users. These include:
- Unblended costs
- Amortized costs
- Blended costs
- Net unblended costs
- Net amortized costs
I will explain the above datasets briefly to give you a rough idea of how AWS provides a well-structured cost.
This is vastly used by most of the AWS users and is presented on the Bills page. It is the default option for analyzing cost using AWS Cost Explorer. It is also the default option for setting custom budgets in AWS Budgets. This represents the usage costs charged to you on a cash basis of accounting. For most users, this is the only dataset they need.
When Unblended costs make little sense, Amortized costs come into the picture. It shows the costs on a rather accrual basis than a cash basis. This cost dataset is most useful for those who have purchased AWS Reservations such as Amazon EC2 Reserved Instances.
If you are using unblended costs as your cost dataset, you might encounter a spike if recurring fees are charged on the first day of a month since savings plans and Reservations often have upfront or recurring monthly fees associated with them.
Amortized costs help distribute these recurring costs evenly across the month. If you seek to gain insight into the effective daily costs associated with your reservation portfolio, amortized cost datasets are a powerful tool.
These were originally created for the users who want to merge their billing under a single paying account. The way these are calculated makes it unlikely for people to use it.
In this dataset, each account’s service usage is multiplied against a blended rate, which is an average rate of on-demand usage, Savings Plans, and Reservation related usage. This is consumed by member accounts in an organization for a specific service.
Other costs dataset
In rare case scenarios, AWS users can take advantage of specialized discounts. The net unblended costs dataset gives the total cost after applying discounts while the net amortized costs dataset adds additional logic to amortize discount related data besides your Savings Plans or Reservation related charges.
One should choose cost datasets carefully. If you are using Savings Plans or Reservations, you will be benefited from Amortized costs otherwise you must choose unblended costs dataset. If you are operating at scale or have a specialized use case, you might go for other cost datasets like net unblended or net amortized datasets.