OrangeGrid Examples

This repository provides code examples for commonly used applications within the OrangeGrid cluster. In addition to exploring these examples on the web you can use git to download them into your home directory on the cluster which will allow you to run them directly. The command is

git clone http://github.com/SyracuseUniversity/OrangeGridExamples

Start here

• hostname: A simple example of submitting a job to the cluster, monitoring it, and checking its output.

Getting the most out of OrangeGrid

These examples discuss general techniques for optimizing performance and throughput

• Checkpointing: Learn how to save work that your jobs are doing, so if they exit for any reason HTCondor can restart them where they left off.
• Parallelism: Learn how to divide a task into lots of smaller tasks that can run independently so they can spread out over the cluster.
• File management: Learn how to arrange your data into chunks that are both more efficient for later analyses and perform better on the cluster.
• multipleJobs: Learn how to run multiple jobs from one submit file.
• HTCondorDiagnostics: Check the status of the cluster.

Using particular languages and libraries

• python: Use Python and Python packages with the Conda package manager.
• uv: Use Python and Python packages with the uv package manager.
• tensorflow: Use the Tensorflow package.
• PyTorch: Use the PyTorch package (note, the uv example also covers this in a way that may be easier to use).
• Ollama: Run an LLM non-interactively, choosing from among numerous models.
• julia: Use the Julia language and its libraries.
• octave: Use Octave, a free and open source math-oriented language that is largely compatibe with Matlab.
• R: Use the R language and its libraries.
• Apptainer: Simplify the installation of complex packages by pulling containers from Dockerhub or building new containers.
• blender: Render images and movies.
• CVMFS: Use software and data distributed through the CERN Virtual File System
• Grobid: Use GROBID, a machine learning library for extracting, parsing and re-structuring raw documents such as PDF into structured documents.

Workflow managers

Single submit files are great for anything from a single job to thousands of jobs where the same command is run on numerous arguments. However research often entails more complex arrangements of jobs, where an initial stage will create files which are needed by a later stage or postprocessing can only be run once a set of analyses have completed. In general there may be arbitrary dependencies between jobs. While it is always possible to manage these manually, waiting for one set of jobs to complete before running the next, it is much more convinient to have a workflow manager handle the dependencies. There are many such systems, suitable for different situations.

• DAGMan (Directed Acyclic Graph Manager) is HTCondor's native workflow manager. Using it involves creating a simple text file describing the jobs and their connections (or writing a program to generate such a file). We discuss DAGMan in the section on Parallelism.

• Pegasus is an extremely powerful, but also fairly complex, workflow management system developed by the Information Sciences Institute at the University of Southern California.

• SnakeMake is a tool for running workflows spficifed by the relationships between input and output files of individual processes.

• NextFlow (coming soon)

Need Help?

Additional how-to documentation, such as connecting to clusters and running jobs, is available in Answers.

If you would like to contact us directly for assistance or requesting access, email researchcomputing@syr.edu.