Job Metrics dashboard

The Job Metrics dashboard is a monitoring and analytics tool that gives Organization Admins detailed insights into how compute jobs are utilizing computational resources across their organization.

The dashboard is designed to help you understand the efficiency of your computational workload by comparing requested resources against actual usage, monitoring success rates, and more. Use it to monitor Platform compute usage, optimize resource allocation, and reduce compute costs.

Key Features

Detailed Metrics View

The dashboard displays both Requested and Used amounts for CPU and memory across all jobs within a selected timeframe. This dual-metric approach helps you identify inefficiencies—if there's a significant gap between requested and used resources, you may be over-provisioning a job and incurring unnecessary costs.

Flexible Date Range Selection

Analyze compute data over any time period from 1 to 31 days, beginning up to 1 year in the past. Choose the date range that best fits your reporting needs, whether you're conducting a daily check-in or a monthly review.

Success Rate Tracking

Monitor job completion rates to identify issues affecting your workflows. The success rate metric helps you spot patterns in job failures and take corrective action. This metric can be a great way to identify jobs that are consistently failing and unschedule or fix them.

Approximate Compute Load

Sort the approximate compute load column to determine which jobs are contributing the most to your cluster’s compute hours. This metric combines resource requests with total runtime to give you a heuristic of the job’s overall compute impact.

CSV Export Functionality

Export dashboard data to CSV format for deeper analysis, custom reporting, or integration with other tools. This allows you to perform advanced analytics or share metrics with stakeholders outside the platform.

Accessing the Job Metrics dashboard

The Job Metrics dashboard is only available to organization admins. To access it:

1. Log in to Civis Platform.
2. Navigate to the Platform Usage Overview via the Admin menu.
3. Select "Historical Compute". Job Metrics will open by default.

Using the dashboard

Viewing Job Metrics

The main dashboard view displays a table of jobs with the following columns:

• Job Name - Name of the job. Links to the job’s details page. Please note that you may need to enable Superadmin Mode in order to visit all links.
• Job ID - Job ID
• Compute User - Name of the user who is responsible for executing the job.  
• Template ID - ID of the job’s backing template script, if applicable.
• Success Rate - Percentage of runs within the timeframe that were successful, calculated as: (Successful Runs / Total Runs) × 100. A healthy success rate is typically above 80%, though acceptable rates vary by use case.
• Run Count - Number of runs within the timeframe. Partial runs, i.e. runs starting before the timeframe or ending after the timeframe, count as 0.5 of a run.
• Total Runtime - The total runtime for all runs within the timeframe. For partial runs, i.e. runs starting before the timeframe or ending after the timeframe, only the runtime within the timeframe will contribute to this sum.
• Approximate Compute Load - This metric can be used to discover which jobs are contributing the most to your cluster’s overall compute load. It compares your job’s average CPU and memory requests relative to the available CPU and memory on an standard instance type to determine which constitutes the larger percentage used, then multiplies that percentage by the job’s total runtime. Please note that this is a relative metric, intended to help you track down your most compute intensive jobs; it does not translate directly to compute hours.
• Average CPU Requested - Average millicores of CPU requested by runs within the timeframe.
• Max CPU Used - Max millicores of CPU used by a run within the timeframe. If a run finishes quickly, e.g. in only a few minutes, this metric may not be available.
• Average Memory Requested -  Average megabytes of memory requested by runs within the timeframe.
• Max Memory Used - Max megabytes of memory used by a run within the timeframe. If a run finishes quickly, e.g. in only a few minutes, this metric may not be available.

Use the sorting functionality to organize jobs by any column, making it easy to identify your highest resource consumers or least efficient jobs.

Sorting & Filtering

Table data can be sorted by hovering over the column you’d like to sort by and clicking the arrow that appears: 

You can also filter the table to show data that matches a certain pattern. E.g. you could filter for jobs with “Production” in their name or jobs belonging to a particular user. Filtering is available from the triple dots action menu that appears beside each column name on hover:

Columns can also be pinned, hidden, or sorted from the column action menu. Columns may be re-arranged via dragging and dropping them.

Selecting a Date Range

1. Click the date range selector at the top of the dashboard
2. Choose a start date and end date (maximum 31-day span)
3. Click the search button to fetch new data for the dashboard

Start dates may be no more than 1 year in the past.

Exporting Data

To export dashboard data:

1. Configure your desired date range and filters, and click search
   • Note: You must load data into the table before it can be exported
2. Click the "Export to CSV" button in the upper right corner
3. The CSV file will download to your local machine

The exported file includes all metrics for further analysis in spreadsheet software or data analysis tools.

Best Practices

• Review metrics regularly: Check the dashboard weekly or monthly to identify trends and optimization opportunities.
• Right-size your jobs: Use the max used vs. average requested metrics to adjust resource allocations and reduce unused compute.
• Monitor success rates: Investigate any jobs with declining success rates to prevent workflow disruptions.
• Share insights with your team: Export data and share findings with job owners to promote resource efficiency across your organization.
• Track improvements over time: Use consistent date ranges to measure the impact of optimization efforts.

FAQs

Who can access the Job Metrics dashboard?

Only organization admins have access to the Job Metrics dashboard. Since the Job Metrics dashboard includes information about all compute jobs within the organization, regardless of individual permissions, access is scoped to only organization admins for security.

How often is the dashboard data updated?

Dashboard data is updated hourly. 

Can I view metrics for a specific user?

Yes, see the “Sorting & Filtering” section above to learn how. 

What's the maximum date range I can analyze?

The dashboard supports date ranges from 1 to 31 days. For analysis spanning longer periods, export multiple date ranges and combine the data externally. Start dates may be up to 1 year in the past.

How can I use this data to reduce costs?

Look for patterns of over-provisioning where jobs consistently use significantly less CPU or memory than requested. Work with job owners to right-size these jobs, which will reduce your overall compute costs without impacting performance.

What should I do if I see consistently low success rates?

Low success rates may indicate issues with job configuration, credentials, resource constraints, or dependencies. Review job logs for specific errors, and consider increasing resources if jobs are failing due to memory or CPU limitations.

For questions or feedback about the Job Metrics dashboard, please reach out to support@civisanalytics.com.