Pay-as-you-go is a billing method where you pay after using resources. It suits scenarios with highly variable or temporary resource demands. Charges are calculated based on actual CU-hour usage and the unit price in your region. - E-MapReduce

Compute resources (pay-as-you-go) use a post-payment billing method. You do not need to purchase large amounts of resources in advance. The system bills you based on your workspace’s actual resource usage.

Important notes

EMR Serverless Spark uses pay-as-you-go billing. During peak periods, resources may be preempted, and timely resource availability cannot be fully guaranteed.

Pay-as-you-go details

Attribute	Description
Scenarios	Pay-as-you-go is ideal for the following scenarios: Your workload varies frequently. Your resource needs are temporary or bursty.
Billing rule	Your bill for each billing cycle (1 hour) equals the resource fee based on actual compute usage. This is calculated by converting total consumed resources in the cycle into CU-hours, then multiplying by the hourly unit price in your region. The compute resource fee is:`CU usage within 1 hour × hourly unit price in your region`. If, within an hour, your workspace’s cumulative CPU-to-memory ratio falls between 1:2 (inclusive) and 1:3 (exclusive), you receive a 20% discount when converting to CU. For example, consider the following Spark job configuration. `spark.driver.cores 1 spark.driver.memory 1g spark.driver.memoryOverhead 1g spark.executor.cores 1 spark.executor.memory 1g spark.executor.memoryOverhead 1g spark.executor.instances 2` This uses `3 cores and 6 GB` of memory per minute. Over one hour, this totals `180 core-minutes and 360 GB-minutes`. Using the formula `max(180/1, 360/4)`, this converts to 180 CU-minutes. Because the CPU-to-memory ratio is 1:2, you qualify for the 20% discount. 180 CU-minutes × 0.8 = 144 CU-minutes, or 2.4 CU-hours. Therefore, the total cost for one hour is `2.4 CU-hours × hourly unit price in your region`. For unit prices by region, see Regional unit prices. Important The above formula applies to a single EMR Serverless Spark workspace. If you have multiple workspaces, your bill includes the total fees for all workspaces. Billing starts when the container runtime begins. Spark Driver and Executor initialization occurs after the container starts running. Therefore, CU-hours calculated from Driver and Executor runtimes shown in the Spark UI may differ from actual billed CU-hours. Refer to your final bill for accurate charges.
Billing cycle	Fees are calculated hourly on the hour (based on UTC+8 time). After calculation, a new billing cycle begins. At the end of each cycle, the system generates a bill and deducts the corresponding amount from your account. Billing records may lag behind actual usage.

Regional unit prices

The tables below list pricing for workspaces in different regions.

Important

Actual prices are shown on the product purchase page.

Mainland China
Region Name
Unit price (USD/CU/hour)
China (Beijing)
China (Shanghai)
China (Hangzhou)
China (Shenzhen)
0.050302
China (Hong Kong)
0.074048
Other regions
Region
Unit price (USD/CU/hour)
Indonesia (Jakarta)
0.067106
Germany (Frankfurt)
0.064792
Singapore
0.067106
US (Virginia)
0.053801
US (Silicon Valley)
0.053801
Japan (Tokyo)
0.069420
Mexico
0.051220

Region	Unit price (USD/CU/hour)
Indonesia (Jakarta)	0.067106
Germany (Frankfurt)	0.064792
Singapore	0.067106
US (Virginia)	0.053801
US (Silicon Valley)	0.053801
Japan (Tokyo)	0.069420
Mexico	0.051220