DynO: Dynamic Onloading of Deep Neural Networks from Cloud to Device

TL;DR

DynO is a novel distributed inference framework that dynamically balances cloud and device resources for CNNs, significantly improving performance and reducing data transfer compared to existing methods.

Contribution

It introduces a CNN-specific data packing method and a dynamic scheduler to optimize inference partitioning and data precision in real-time.

Findings

Over 10x throughput improvement over device-only execution

Up to 7.9x performance gain over CNN offloading systems

Up to 60x reduction in data transferred

Abstract

Recently, there has been an explosive growth of mobile and embedded applications using convolutional neural networks(CNNs). To alleviate their excessive computational demands, developers have traditionally resorted to cloud offloading, inducing high infrastructure costs and a strong dependence on networking conditions. On the other end, the emergence of powerful SoCs is gradually enabling on-device execution. Nonetheless, low- and mid-tier platforms still struggle to run state-of-the-art CNNs sufficiently. In this paper, we present DynO, a distributed inference framework that combines the best of both worlds to address several challenges, such as device heterogeneity, varying bandwidth and multi-objective requirements. Key components that enable this are its novel CNN-specific data packing method, which exploits the variability of precision needs in different parts of the CNN when onloading computation, and its novel scheduler that jointly tunes the partition point and transferred data precision at run time to adapt inference to its execution environment. Quantitative evaluation shows that DynO outperforms the current state-of-the-art, improving throughput by over an order of magnitude over device-only execution and up to 7.9x over competing CNN offloading systems, with up to 60x less data transferred.