Performance Reports

Overview

Dask and Coiled help you understand computational performance and profile your distributed workload in terms of cluster utilization, workload profiling, network communication, task execution, and more. Dask provides an interactive dashboard that shows various plots and tables that update with live information as computations are running.

In addition to the interactive dashboard, Dask has the ability to generate static performance reports and save the results to a standalone HTML file. These performance reports are useful for later inspection since they can be viewed after your computation is finished and your Coiled cluster is no longer running.

../_images/performance-report-profile.png

Coiled also provides functionality to generate performance reports using the same functionality in Dask, but with the added step of uploading the performance report to Coiled Cloud. This makes it easier to share performance reports with other members of your team or with Dask experts at Coiled without having to email or send around an HTML file.

Generating performance reports

To generate a performance report and upload it to Coiled Cloud, simply wrap the Dask code that you want to profile with the coiled.performance_report() context manager:

from coiled import performance_report

with performance_report(filename="dask-report.html"):
    df.groupby(...).value.mean().compute()  ## Your dask computation(s)

After the computation finishes, the Coiled client will output a message that includes a link to your hosted performance report:

Performance Report Available at: https://cloud.coiled.io/your-username/reports/74

That’s it! Your performance report is now available on Coiled Cloud and is ready to be viewed or shared with others. You can open the link in your browser to view and explore it within the Coiled Cloud interface:

../_images/performance-report-tasks.png

The video below walks you through generating a performance report using Coiled.