Weichen Chen | Searcharxiv

arXiv2025

Taiji: A DPU Memory Elasticity Solution for In-production Cloud Environments

The growth of cloud computing drives data centers toward higher density and efficiency. Data processing units (DPUs) enhance server network and storage performance but face challenges such as long hardware upgrade cycles and limited resources. To address these, we propose Taiji, a resource-elasticity architecture for DPUs. Combining hybrid virtualization with parallel memory swapping, Taiji switches the DPU's operating system (OS) into a guest OS and inserts a lightweight virtualization layer, making nearly all DPU memory swappable. It achieves memory overcommitment for the switched guest OS via high-performance memory elasticity, fully transparent to upper-layer applications, and supports hot-switch and hot-upgrade to meet in-production cloud requirements. Experiments show that Taiji expands DPU memory resources by over 50%, maintains virtualization overhead around 5%, and ensures 90% of swap-ins complete within 10 microseconds. Taiji delivers an efficient, reliable, low-overhead elasticity solution for DPUs and is deployed in large-scale production systems across more than 30,000 servers.

cs.OS↗

arXiv2021

Pedestrian Attribute Recognition in Video Surveillance Scenarios Based on View-attribute Attention Localization

Pedestrian attribute recognition in surveillance scenarios is still a challenging task due to the inaccurate localization of specific attributes. In this paper, we propose a novel view-attribute localization method based on attention (VALA), which utilizes view information to guide the recognition process to focus on specific attributes and attention mechanism to localize specific attribute-corresponding areas. Concretely, view information is leveraged by the view prediction branch to generate four view weights that represent the confidences for attributes from different views. View weights are then delivered back to compose specific view-attributes, which will participate and supervise deep feature extraction. In order to explore the spatial location of a view-attribute, regional attention is introduced to aggregate spatial information and encode inter-channel dependencies of the view feature. Subsequently, a fine attentive attribute-specific region is localized, and regional weights for the view-attribute from different spatial locations are gained by the regional attention. The final view-attribute recognition outcome is obtained by combining the view weights with the regional weights. Experiments on three wide datasets (RAP, RAPv2, and PA-100K) demonstrate the effectiveness of our approach compared with state-of-the-art methods.

cs.CV↗