Project Grant 2153502

This National Science Foundation Office of Advanced Cyberinfrastructure Project Grant of $190,926 will support the development of a novel serverless computing platform with a hybrid CPU/GPU architecture to accelerate deep learning application development and deployment. Funded under the Computer and Information Science and Engineering program (CFDA 47.070), key products include a shim virtualized GPU layer and refactored container subsystem to enable high-performance GPU sharing for concurrent serverless functions. The shim layer will provide fine-grained GPU resource provisioning and isolation through API remoting, while optimizing performance via caching and scheduling. The container subsystem will accelerate the deep learning lifecycle by exploiting model structures to parallelize CPU-to-GPU memory copying and model execution. Louisiana State University will receive the full award amount to develop these technologies from May 1, 2022 to April 30, 2024.