This topic describes how to select a cost-effective billing method.
Billing methods
MaxCompute supports the following billing methods for computing, storage, and download resources:
Subscription: Computing resources are charged on a monthly or annual basis. Storage and download resources are charged on a pay-as-you-go basis.
Pay-as-you-go: Storage, computing, and download resources are all charged on a pay-as-you-go basis.
Hourly billing: In addition to subscription computing resources, extra computing resources are charged on an hourly basis. You can scale out or in computing resources based on your business requirements. You are charged for the temporarily added computing resources based on the number of compute units (CUs) that are added and the usage duration.
For more information, see Overview. You can select a billing method with the help of Total Cost of Ownership (TCO) tools and the best practices of cost estimation.
TCO tools
You can use the following TCO tools to estimate costs:
MaxCompute price calculator: This tool is suitable for the subscription billing method. To calculate the monthly cost, enter the required computing resources and the amount of the data that you want to upload and download to the MaxCompute price calculator.
CostSQL: This tool is suitable for the pay-as-you-go billing method.
You can run the COST SQL command to estimate the cost of an SQL job before you execute the SQL job in a production environment. For more information, see Cost estimation.
If you use IntelliJ IDEA, the system automatically estimate costs when you submit SQL scripts. For more information, see Develop and submit an SQL script.
If you use DataWorks, you can also estimate costs.
NoteThe costs of some SQL jobs cannot be estimated, such as SQL jobs that involve external tables.
The actual costs are subject to final bills.
Best practices of cost estimation
This section provides some cost estimation examples and tips for your reference. You can select a cost-effective billing method based on the information.
Billing methods for 1 TB of data
The following table describes estimated costs for reference.
Billing method
Business scenario
Response speed
Estimated cost per month
Subscription
Compute-intensive
Within a few minutes
USD 3,768
Storage-intensive
Within a few hours
USD 1,177.5
Pay-as-you-go
USD 1,413(The cost is estimated based on the SQL complexity of 1 and the execution frequency of once per day.)
If you select the subscription billing method, the costs vary depending on your business type:
Compute-intensive scenario: In this scenario, a large number of CPU resources are required. 160 CUs are used to process 1 TB of data. The system responds to a request within a few minutes. The estimated cost is USD 3,768 per month.
Storage-intensive scenario: If your jobs are not sensitive to the response speed, we recommend that you purchase a subscription storage plan. About 50 CUs are used to process 1 TB of data. The system responds to a request within a few hours. The estimated cost is USD 1,177.5 per month.
If you select the pay-as-you-go billing method, the estimated cost for processing 1 TB of data with the SQL complexity of 1 is USD 47.1 per day and USD 1,413 per month. If the data is processed multiple times per day, the cost is multiplied.
If you migrate data to the cloud for the first time, we recommend that you select the pay-as-you-go billing method first. Perform a Proof of Concept (POC) test to calculate the approximate number of CU-hours used for your jobs. Then, calculate the number of CUs that you need to purchase based on the number of CU-hours. You can calculate the number of CUs based on TASKS_HISTORY of the Information Schema service.
Billing methods for Hadoop users to migrate data to the cloud
For example, a Hadoop cluster has one controller node and five compute nodes. Each node has 32 cores, equivalent to 32 CPUs. The total number of CPUs for the compute nodes is 160. The estimated cost of the cluster is USD 3,768 per month with no discounts or promotional offers applied.
MaxCompute does not require any controller nodes. The performance of MaxCompute is 80% higher than the performance of Hive. MaxCompute frees you from operations and maintenance (O&M) and reduces costs.
Mixed billing methods
Subscription billing method for production businesses, such as hourly extract, transform, load (ETL), and pay-as-you-go billing method for aperiodic jobs or ad hoc queries
We recommend that you select the subscription billing method for periodic computing jobs that are frequently executed and the pay-as-you-go billing method for aperiodic jobs that are used to process large amounts of data. In pay-as-you-go mode, you can choose not to store data. Instead, you can read data from tables under other accounts. This reduces data storage costs. Authorization is required for computing operations on tables under different accounts. For more information, see Create a project-level role.
Subscription billing method for aperiodic jobs or ad hoc queries and pay-as-you-go billing method for production businesses, such as daily ETL
Daily data testing may cause the issue of uncontrollable costs. To avoid this issue, you can add data testing and aperiodic jobs to fixed resource groups. Then, use MaxCompute Management to configure custom development groups and business intelligence (BI) groups. If production jobs are executed only once per day, you can add them to a pay-as-you-go resource group.
Switching between billing methods
If you select the subscription billing method, you can upgrade or degrade the configurations in the following scenarios: The data volume changes, and the purchased resources are insufficient or become idle. For more information, see Upgrade or downgrade configurations.
You can also switch between the billing methods. For more information, see Switch billing methods.
Before you switch the billing method to subscription, evaluate the computing performance and cycles of jobs to determine the number of CUs you need to purchase. If the CUs you purchase are insufficient, the computing cycle of a job may be prolonged, and the computing performance may not meet your expectations. If this occurs, you may need to switch the billing method again.