Giving a talk at our open community calls is a great way for community members to share progress and updates on their Parsl-related projects/work. In doing so, users can work through challenges together, answer questions, and raise awareness of their work. If you are interested in giving a talk at one of our upcoming calls, please sign up for a slot in our 2024 schedule.
Check out recorded talks from past open community calls here ⬇ or on the Parsl YouTube channel.
Recording Watch Link & Details | Overview, Notes & Links |
---|---|
![]() Speaker: Sander Vandenhaute |
In this talk, Sander will discuss psiflow, a scalable molecular simulation framework that targets non-Parsl-savvy users. He’ll discuss the easy setup of Parsl providers and executors on clusters, combining containers and parallel efficiency, setting ulimits, and smarter Futures. |
![]() Speaker: Ben Clifford |
In the past month, a few questions have come up from different groups about dealing with task outcomes more than the default of passing on successful results and failing on exceptions when using Parsl, such as "How can I tell Parsl that some of its checkpointed results are actually wrong and that some failures should be checkpointed because that error means the task will never run successfully?" "How can I pass on an exception to the next task rather than abort the whole pipeline?" In this talk, Ben Clifford will discuss how dealing with this might fit into Parsl. |
![]() Speaker: Ben Clifford |
In this talk, Ben discusses his work with monitoring radio plugins. |
![]() Speaker: Logan Ward |
This talk is meant to elicit a discussion on how our community defines the applications run by Parsl. For example, Logan rarely employs the decorators prominent in the Parsl documentation and feels strongly about his decision. |
![]() Speaker: Ben Clifford |
In this talk, Ben Clifford will provide a brief history of workflow-related file staging and what the future might hold for Parsl. |
![]() Speaker: Colin Thomas |
The TaskVine executor, developed by Notre Dame’s Cooperative Computing Lab, allows Parsl workflows to utilize the local storage of worker nodes, improving cluster bandwidth and minimizing data movement. This talk will introduce work on extending TaskVine intermediate data handling to Parsl with additional related features. |
![]() Speaker: Ben Clifford |
Covered topics include: Outreachy overview and progress, scaling code and challenges, tidying stderr/stdout, explicit cleanup at exit, flux, MPI discussion, worker drain feature, DESC work, and Debian packaging. |
![]() Speaker: Nishchay Karle |
By leveraging the Common Workflow Language (CWL) and Parsl's parallel scripting capabilities, CWLApp simplifies the execution of command line tasks. |
![]() Speaker: Ben Clifford |
We have several MPI executor implementations now, and they all interact awkwardly, one way or another, with the core Parsl task model. In this talk, Ben discusses the model, how it conflicts with MPI tasks, and some possibilities for sorting it out. |
![]() Speaker: Ben Clifford |
In this talk, Ben Clifford discusses the scale_in functionality work he is developing for Parsl. He provides an overview of the historical and cultural artifacts and events that led to this work, challenges, improvements, and progress. |
![]() Speaker: Reid Mello |
In this talk, Reid will discuss a recent contribution he made to Parsl that allows users to enable encryption for the HighThroughputExecutor by setting its encrypted initialization argument to True . Under the hood, we use CurveZMQ to encrypt all communication channels between the executor and related nodes. |