E-MapReduce:Billing Items and Billing Methods

Computing Resources

You are billed for computing resources based on either your subscription plan or your actual pay-as-you-go usage. The unit of measurement is CU.

1 CU = 1 vCPU + 4 GiB memory. This is the basic unit of computing power for an EMR Serverless Spark workspace. CU usage for a computing task depends on the volume of data processed, computational complexity, data distribution, and whether you enable the Fusion engine. Enabling the Fusion engine does not increase resource costs. It typically reduces job runtime by more than 30%.

If your vCPU-to-memory ratio is less than 1:4, see EMR Serverless Spark CU promotions for details on the conversion.

The following table shows the processing capacity of 1 CU.

Model Calls

You are billed for model calls based on your actual usage of built-in AI Center model services. The unit of measurement is thousands of Tokens.

Token usage is estimated as follows: • 1 Chinese character ≈ 1.5–2 Tokens • 1 English letter ≈ 0.25 Tokens • 1 English word ≈ 1.3 Tokens

• Alibaba Cloud Model Studio: ≈ 4–5 Tokens

• Hello World: ≈ 2 Tokens

For example, in a data masking scenario, you process 1,000 text records in a batch. The following is an example SQL statement:

select
  ai_query (
    concat(
      'Mask the following text according to these rules:
      1. Chinese names: Detect all Chinese names (2–4 characters). Keep the first and last characters. Replace middle characters with `*`.
      2. Mobile phone numbers: Detect Chinese mainland mobile numbers (11 digits, starting with 1). Keep the first 3 and last 4 digits. Replace the middle 4 digits with `****`.
      3. Physical addresses: Detect specific physical address information, such as province, city, district, street, and building number. Keep only the province, city, and district (county) levels. Replace all detailed street, building number, and community names after the district/county level with `***`.
      Output requirements:
      Return only the following three lines. Do not include explanations, original text, or other content:
      Name: [result]
      Mobile: [result]
      Address: [result]',
      user_info
    )
  )AS masked_text
from
  user_tbl
;

For example, a user_info value can be: “Mr. Li Si lives at No. xxx, Moumou Road, Zhangjiang Hi-Tech Park, Pudong New Area, Shanghai. His mobile number is 159********.”

The estimated token usage for one record is as follows:

The estimated Token usage for 10,000 records is as follows: