Pay-as-you-go suits workloads with variable job volumes or unpredictable storage needs. You pay only for what you use — no upfront commitment required.
Resources
Pay-as-you-go covers four resource types. Computing and Internet downloads are billed; storage is billed on compressed size; uploads are free.
| Resource | Billing basis | Details |
|---|---|---|
| Computing resources | Compute units (CUs) consumed per job | Shared resource pool. Supports SQL (with user-defined functions (UDFs)), MapReduce, Spark, Lightning, and Graph jobs. High-priority jobs can preempt low-priority jobs. You cannot specify or limit the amount of resources that computing jobs can consume. See Computing pricing. |
| Storage resources | Compressed size of data stored in tables | MaxCompute compresses data before storage — typically to about 20% of the original size. See Storage pricing (pay-as-you-go). |
| Data uploads | Not billed | Uploading data to MaxCompute is free. |
| Data downloads | Internet downloads only | Downloads over the Internet are billed. See Download pricing (pay-as-you-go). |
Usage notes
-
To activate pay-as-you-go, see Activate MaxCompute and DataWorks.
-
After activation, projects in the same region can use pay-as-you-go computing resources. There is no limit on the number of jobs.
-
Computing resources come from a shared resource pool. If the pool is fully consumed when a job runs, the job waits for resources to become available or preempts lower-priority jobs — which can increase execution time. For example, a job that requires 1,200 CUs runs immediately if the pool has enough capacity. If all the resources in the resource pool are consumed or the number of available CUs is less than 1,200, the job must wait for resource release or preempt the remaining resources, which may delay completion.
When many jobs run simultaneously, multi-user resource contention can cause response latency. Plan accordingly for time-sensitive workloads.
-
To switch to a different billing method, see Switch between billing methods.
What's next
For non-time-sensitive jobs on large volumes of low-value data — such as user behavior logs or system logs — consider Pay-as-you-go Spot Edition to reduce costs further.