Hi Apache Airavata community, I wanted to share the experiment I conducted on CPU/GPU checkpointing using CRIU and the CRIUgpu plugin, along with the results I collected.
After completing the initial research on GPU checkpointing, I created a standalone experiment environment to validate CRIU and CRIUgpu behavior before integrating checkpoint/restore functionality into CyberShuttle Linkspan. The purpose of this stage was to confirm that a running computational workload could be checkpointed and restored correctly, understand the practical requirements of CRIU and the NVIDIA CUDA plugin, measure representative checkpoint/restore behavior, and identify implementation issues that would need to be addressed during the later Linkspan integration. The experiment setup *repository *is available here: "https://github.com/Rosette2237/criu-experiment"; The experiment *results and logs* are available here: " https://drive.google.com/drive/folders/1cWpKFsCFs063Q8emCKqsbM7QKjQqpab4?usp=sharing " I also posted an approximately *8-minute-read blog on the Apache Airavata Wiki* ( https://cwiki.apache.org/confluence/spaces/AIRAVATA/pages/449285378/Checkpointing+Experiments) that describes the experiment in more detail, from the experiment design through the results. If you are interested, please feel free to take a look at the links I shared. I would also be happy to discuss any questions or feedback you may have. Best regards, Boyang Gong
