Computation pricing

Subscription (CU cost)

Payment by subscription is only available on the Alibaba Cloud DTPlus Platform . Subscription allows you to pay an initial fee (monthly or annually) for your entire reserved resources. The basic unit of such resources is defined as CU (Compute Unit). One CU includes 1 core CPU and 4 GB of memory.

We recommend that new users use the Pay-As-You-Go billing method, because this allows you to gauge your resource usage without unnecessary costs. Payment by subscription is only available on the Alibaba Cloud DTPlus platform.

Pay-As-You-Go for SQL standard tasks

SQL tasks are paid after volume, that is, SQL is charged after I/O:

Every SQL task is billed according to Data Input Size and SQL Complexity. Once the SQL task is completed, MaxCompute sends its metering information to the billing system, which calculates the fee and adds it to the next payment.

The MaxCompute SQL task is charged according to I/O for each job. All daily measurement information is paid next day.

The bill for SQL tasks is calculated as follows.

Computing Cost of One SQL Task = Data Input Size × SQL Complexity × SQL Price

The price is as follows.

• Data Input Size:The actual size of the data that an SQL statement scans. Most SQL statements have partition filtering and column pruning, so this value is generally less than the source table data size.
  • Column pruning: For example, the submitted SQL is select f1,f2,f3 from t1; Only the data size of three columns (f1, f2, and f3) in t1 are charged.
  • Partition filtering: For example, a SQL statement includes ds>”20130101”. The “ds” is a partition column. The data size is calculated only according to the read partition, rather than the data of other partitions, and then billed.
• SQL Complexity:First, MaxCompute counts keywords in SQL statements, and then converts to SQL complexity.
  • SQL keyword number = join Number + group by number + order by number + distinct number + window function number + max (insert into Number -1, 1)
  • SQL complexity calculation:
    • If SQL keyword number is less than or equal to 3, the complexity is 1.
    • If SQL keyword number is less than or equal to 6, the complexity is 1.5.
    • If SQL keyword number is less than or equal to 19, the complexity is 2.
    • If SQL keyword number is greater than or equal to 20, the complexity is 4.

The input SQL statement for calculating SQL Complexity is as follows:

cost sql <SQL Sentence>;

An example of a SQL statement is as follows:

odps@ $odps_project >cost sql SELECT DISTINCT total1 FROM
(SELECT id1, COUNT(f1) AS total1 FROM in1 GROUP BY id1) tmp1
ORDER BY total1 DESC LIMIT 100;
Intput:1825361100.8 Bytes
Complexity:1.5

The preceding SQL includes 4 keywords (one DISTINCT, one COUNT, one GROUP BY, and one ORDER), so the SQL complexity is 1.5. If the data volume of table “in1” is 1.7 GB, then the actual consumption is as follows:

1.7 * 1.5 * 0.0438 = 0.11 USD

Pay-As-You-Go for SQL external tables tasks

The billing for MaxCompute SQL external tables commenced on 24th July, 2019. The billing for one external table SQL task is calculated as follows.

Computing Cost of One SQL Task = Data Input Size × SQL Price

The price for an SQL task is 0.0044 USD/GB/Complexity. The complexity coefficient is 1. All metering information is calculated and summarized by the end of the day and billing is then generated for the next day.

Pay-As-You-Go for MapReduce

In December 19, 2017, MaxCompute began charging for MapReduce (MR) tasks. The billing of an MR task is calculated as follows:

Computation Cost of One MR task = Total Time × MR Price (USD)

The calculating time for each successful MR task = Execution time (hours) × Number of cores that task calls.

If one MR task calls 100 cores, and the task takes 30 minutes to complete, the calculating time for the MR task is: 0.5 hours × 100 cores = 50 hours.

After the MR task is finished, MaxCompute sends its metering information to the billing system, which calculates the fee and adds it to the next payment.

Pay-As-You-Go for Spark

Overall computation cost for all Spark on MaxCompute tasks on a day = Total number of compute hours for all tasks on the day × Unit price (0.1041 USD/Hour/Task)

Spark Task Compute hours=Max [Number of CPU cores consumed x Computation duration,roundup (Memory space consumed x Computation duration/4)]

Max[2 × 1, roundup (5 × 1/4)] = 2

Max [2 × 1, roundup (10 × 1/4)] = 3