Litcius/Paper detail

Automated Runtime-Aware Scheduling for Multi-Tenant DNN Inference on GPU

Fuxun Yu, Shawn Bray, Di Wang, Longfei Shangguan, Xulong Tang, Chenchen Liu, Xiang Chen

20212021 IEEE/ACM International Conference On Computer Aided Design (ICCAD)45 citationsDOIOpen Access PDF

Abstract

With the fast development of deep neural networks (DNNs), many real-world applications are adopting multiple models to conduct compound tasks, such as co-running classification, detection, and segmentation models on autonomous vehicles. Such multi-tenant DNN inference cases greatly exacerbate the computational complexity and call for comprehensive collaboration for graph-level operator scheduling, runtime-level resource awareness, as well as hardware scheduler support. However, the current scheduling support for such multi-tenant inference is still relatively backward. In this work, we propose a resource-aware scheduling framework for efficient multi-tenant DNN inference on GPU, which automatically coordinates DNN computing in different execution levels. Leveraging the unified scheduling intermediate representation and the automated ML-based searching algorithm, optimal schedules could be generated to wisely adjust model concurrency and interleave DNN model operators, maintaining a continuously balanced resource utilization across the entire inference process, and eventually improving the runtime efficiency. Experiments show that we could consistently achieve <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$1.3\times\sim 1.7\times$</tex> speed-up, comparing to regular DNN runtime libraries (e.g., CuDNN, TVM) and particular concurrent scheduling methods (e.g., NVIDIA Multi-Stream).

Topics & Concepts

Computer scienceInferenceScheduling (production processes)Distributed computingConcurrencyArtificial intelligenceParallel computingEconomicsOperations managementAdvanced Neural Network ApplicationsMachine Learning and Data ClassificationAdversarial Robustness in Machine Learning
Automated Runtime-Aware Scheduling for Multi-Tenant DNN Inference on GPU | Litcius