Job Insights-MaxCompute(MaxCompute)-阿里云帮助中心

MaxCompute provides the Job Insights feature, which allows you to view job summaries, intelligent diagnostics details, resource consumption, and the resource allocation of a computing quota at a specific time. You can also compare historical run metrics for similar jobs, which are identified by the same signature or ExtNodeId. Use this data to optimize job execution and improve efficiency.

Accessing job insights

Log on to the MaxCompute console and select a region in the upper-left corner.
In the left-side navigation pane, choose Workspace > Jobs.
In the job list, locate the desired instance ID and click Insights in the Actions column to open the Job Insights page.

Switch between the tabs to get the information you need. The tabs provide the following features:

Tab	Description
Job Summary	View basic job information, runtime parameters, and intelligent diagnostics details. You can also: Diagnose: Click Diagnose in the upper-right corner of the Job Summary tab to trigger real-time intelligent diagnostics for SQL or SQLRT jobs. LogView: Click LogView in the upper-right corner of the Job Summary tab to open the LogView page and view detailed job execution information. For more information about LogView, see Use Logview 2.0 to view job run information.
Resource Consumption	View job-level resource consumption and the resource allocation of the computing quota at a specific time.
Similar Jobs	Compare metrics such as execution duration, scan size, and resource consumption for historical runs of similar jobs. Jobs are identified as similar based on the same signature or ExtNodeId.

Basic job information and runtime parameters

On the Job Summary tab of the Job Insights page, you can view the basic information and runtime parameters for a job. The following tables describe these parameters:

Basic information

Parameter	Description
Job type	The job type. Valid values: SQL: An SQL job. SQLRT: A query-accelerated SQL job. LOT: A MapReduce job. CUPID: A Spark or Mars job. Algo_Task: A machine learning job. GRAPH: A graph computing job.
Job owner	The job owner.
Job priority	The priority of the job. For more information, see Job priority.
Billing method	The billing method for the job. Valid values: Subscription: The job runs on a subscription computing quota and does not incur postpaid job fees. pay-as-you-go: The job runs on a pay-as-you-go computing quota and generates postpaid job bills.
Submission time	The time when the job was submitted.
Start time	The time when the job acquired its first computing resource. For jobs with very short execution times or those that do not consume computing resources, such as DDL statements, this time is the same as the submission time.
Job status	The job's current status. Valid values: Running: The job is currently executing. Any job that has not completed is in this state. Success: The job ran successfully. Failed: The job failed. Cancelled: The job was canceled. Submitted: The job has been submitted and is waiting for computing resources. Note This status reflects the overall state of the job. However, a job may involve multiple concurrent processes, each with its own sub-status. To view these details, use LogView. For more information, see Use Logview 2.0 to view job run information.
Wait time	The duration from the submission time to the start time.
Execution duration	The duration from the start time to the end time.
End time	The time when the job completed execution.
Total run time	The total duration from the submission time to the end time.

Runtime parameters

Parameter	Description
Project	The project where the job was initiated.
Computing quota	The computing quota used by the job.

Intelligent diagnostic information

When you open the Job Insights page, it automatically triggers a real-time diagnosis for SQL and SQLRT jobs. The Job Summary tab displays detailed Job Summary results and optimization recommendations.

Note

Currently, the intelligent diagnostics feature is available only for SQL and SQLRT jobs. For other job types, the diagnosis option and results are not displayed.

You can also click Diagnose in the upper-right corner of the page at any time to get the latest diagnostic results.

For more information about the diagnostic results, see Job intelligent diagnostics.

Job-level resource consumption

You can view the resource consumption of a job during its lifecycle on the Resource Consumption tab of the Job Insights page.

Chart

Description

Resource usage during the job lifecycle

Displays a time-series chart of computing unit (CU) usage and wait times at the job level. It also shows the trend of CU usage and the number of jobs waiting for CUs at the computing quota level.

If the job-level CU usage is low while the quota-level CU usage is high or consistently at its limit, this indicates that the computing quota is under pressure and other jobs are competing for computing resources with the current job.

Resource allocation of the computing quota at a specific time

Click a point on the chart's time axis to view the computing quota's resource allocation at that moment. The details include the number and priority distribution of all running and waiting jobs.
Click a colored block corresponding to a specific priority to open a list of those jobs and view their details. This helps identify which jobs are competing for computing resources with the current job.
Based on your business needs, you can optimize task execution by adjusting job priorities or managing computing resources. For more information, see Job priority or Manage computing resources (quotas).

For best practices on job-level resource analysis, see Best practices for job-level resource analysis.

Similar jobs

On the Similar Jobs tab of the Job Insights page, you can compare the historical performance metrics of the current job with similar ones.

MaxCompute provides two criteria for identifying similar jobs. The default is Same signature.

Same signature: Jobs with the same SQL job signature as the current job. Available only for SQL jobs.
Same ExtNodeId: Jobs with the same source-side job ID as the current job, such as a node ID from DataWorks. For more information about DataWorks node IDs, see Configure basic properties.

Select a time range of 1d, 7d, or 14d. The default is the time range selected on the Jobs page.

The following performance metrics are available for visual comparison:

Metric	Description
Wait time	The duration from the submission time to the start time.
Execution duration	The duration from the start time to the end time.
CU-hour	The CU-hours consumed by the job. The formula for calculating CU-hours is `MAX(CPU hours, CEIL(Memory hours / 4))`. One CPU-hour represents the use of one CPU core for one hour. `CPU hours = Number of CPU cores × Duration in hours`. One memory-hour represents the use of 1 GB of memory for one hour. `Memory hours = Memory size in GB × Duration in hours`.
Scan size	The amount of data scanned by the job.

The list of similar jobs also includes the following fields for more details:

Field	Description
Instance ID	The instance ID of a similar job. Click the ID to open the Job Insights page for that job.
Submission time	The time when the job was submitted.
Job type	The type of the job.
Job priority	The priority of the job.
Initiating project	The project that initiated the job.
Computing quota	The name of the computing quota used to run the job.
Job owner	The owner of the job.