Search

Deep Neural Network (DNN) is a critical component of a wide range of applications. However, with the rapid growth of the training dataset and model size, communication becomes the bottleneck, resulting in low utilization of computing resources. To accelerate communication, recent works propose to aggregate gradients from multiple workers in the programmable switch to reduce the volume of exchanged data. Unfortunately, since using synchronization transmission to aggregate data, current in-network aggregation designs suffer from the straggler problem, which often occurs in shared clusters due to resource contention. To address this issue, we propose a straggler-aware aggregation transport protocol (SA-ATP), which enables the leading worker to leverage the spare computing and storage resources to help the straggling worker. We implement SA-ATP atop clusters using P4-programmable switches. The evaluation results show that SA-ATP reduces the iteration time by up to 57% and accelerates training by up to $1.8\times $ in real-world benchmark models.

In light of the brisk tempo characterizing contemporary lifestyles and the escalating burden of diverse stressors, the decline in the quality of individuals' sleep has emerged as a consequential issue exerting a notable impact on human physiological health. This article introduces the EdgeGAN system, which proposes a hybrid architecture for medical smart beds aimed at proficiently monitoring sleep quality. The EdgeGAN system seamlessly integrates the Internet of Things (IoT) and edge computing through the incorporation of lightweight Generative Adversarial Networks (GAN) into edge computing devices. The amalgamation of this integration serves to enhance the efficacy of sleep quality monitoring. Relative to conventional sleep monitoring systems, the EdgeGAN system offers reduced computational complexity and streamlined user operation. Furthermore, it adeptly captures long-term temporal dependencies in sleep data, thereby extending the retention time of historical information. It also exhibits exceptional compatibility with sleep monitoring devices. Moreover, the EdgeGAN system possesses the capability to intelligently determine whether to upload pertinent data to the cloud based on user preferences, thereby diminishing reliance on cloud resources. In comparison to traditional cloud platform systems, the EdgeGAN system proposed in this article has the capability to circumvent data blockages arising from increased user requests. This innovation enhances real-time performance and compatibility in sleep monitoring, prioritizing user privacy protection. As a result, it offers an intelligent and convenient solution for the development of future smart medical devices.