Platform For AI:Billing for Machine Learning Designer

Overview

Machine Learning Designer algorithm components may incur costs for PAI algorithm components, PAI-DLC, or MaxCompute. This topic covers PAI algorithm component costs. For other costs, see DLC billing and MaxCompute billing. Prices shown are for reference only.

Billable items

Calculation

Algorithm components have different unit prices by type. Calculate costs as follows:

• Formula: Bill = Billable hours × Unit price

  Billable hours = max(CPU cores × Duration (h), Memory (GB) × Duration (h) / 4)

• Duration: Billing starts when component starts and ends when component stops.

• Example: A data analytics algorithm uses 2 CPU cores and 5 GB memory for 1.5 hours. Billable hours = max(2 × 1.5, 5 × 1.5 / 4) = 3. Bill = 3 × Unit price = 3 × 0.21 = 0.63 USD. Unit prices by algorithm type:

  Algorithm type	Unit price (USD/billable hour)	Algorithm type description
  Data pre-processing (data_manipulation)	0.16	Data pre-processing and feature engineering components.
  Data analytics (data_analysis)	0.21	Statistical analysis, machine learning, time series, network analysis, and finance components. Common algorithms: binning and swing_rec_ext.
  Text analytics (text_analysis)	0.27	Text analytics components.
  Deep learning algorithm (deep_learning)	0.16	Deep learning components such as the CPU version of the TensorFlow framework and EasyRec series algorithms.

  Note

  • If MaxCompute is enabled when activating PAI, MaxCompute resources incur charges when using SQL Script, Join, Union, and Filtering and Mapping components. For details, see MaxCompute billing.

  • If Flink is enabled when activating PAI and selected as compute resource, Flink resources incur charges when using Alink components. For details, see Flink billing.

Billing example

An experiment typically contains multiple algorithm components, each with multiple subtasks. To calculate experiment cost, calculate subtask costs for each component and sum them.

1. Identify algorithm component category.

   1. Log on to the PAI console.

   2. In the upper-left corner, select a region.

   3. In left navigation pane, click Workspaces. On resulting page, click target workspace name.

   4. In navigation pane, choose Model Development and Training > Machine Learning Designer.

   5. On Workflows page, select target workflow and click Go to Workflow.

   6. In component list, find PLDA component. It belongs to text analytics category and is priced at 0.27 USD per billable hour.

      

2. View resources consumed by job.

   1. On experiment canvas, right-click PLDA component.

   2. From the shortcut menu, select View Log.

   3. On Log-PLDA page, each blue link corresponds to a subtask. Click a link.

      

   4. On the LogView page, click the SourceXML tab.

   5. Under TaskPlan, view CPU and Memory.

      • Divide CPU value by 100 to calculate CPU cores. This job uses 1 CPU core.

      • Memory unit is MB. This job uses 1 GB memory.

   6. On the LogView page, click the Job Details tab.

   7. Click task object on AlgoTask_0_0 tab. In dialog box that appears, click Terminated tab. Latency shows runtime of each job.

      This subtask has 49 jobs, and each job runs for about 26 seconds.

3. Calculate subtask cost.

   1. Billable hours for subtask = max(CPU cores × Duration (h), Memory (GB) × Duration (h) / 4) = max(49 × 1 × (26/3600), 49 × 1 × (26/3600) / 4) ≈ 0.35 billable hours

   2. Subtask cost = Billable hours for subtask × Unit price = 0.35 × 0.27 = 0.095 USD

4. Calculate costs of all subtasks in PLDA component and sum them to obtain total component cost.

5. Repeat preceding steps to calculate costs of all components in experiment and sum them to obtain total experiment cost.