Cloud providers commonly employ oversubscription strategies to maximize profitability, leveraging the significant gap between the resources purchased by tenants and those actually consumed by their workloads. However, the temporal volatility of workloads may lead to overload on oversubscribed nodes. To address this issue, existing works typically focus on designing reactive rescheduling mechanisms triggered by overload events or adopt conservative oversubscription strategies to mitigate overload risks. Nonetheless, these solutions compromise either tenant experience or provider profitability. In fact, reducing the temporal volatility of workloads is key to addressing the above challenges. We observe that many workloads exhibit temporal complementarity. Aggregating such workloads can effectively mitigate temporal volatility, thereby improving overall resource utilization. Motivated by this insight, we first design a new metric, called Maximum-based Coefficient of Variation (MCV), to quantify the temporal volatility of workloads. We then propose Hestia, a framework that achieves long-term stable oversubscription through workload aggregation. Specifically, we propose a smoothing-based method to classify workloads suitable for aggregation according to their periodicity. Subsequently, we design an aggregation algorithm to minimize the overall MCV, and treat the aggregated workloads as the units for oversubscription. Experimental results show that, using CPU as a representative example, Hestia reduces MCV by 43.3% and increases oversubscription profit by 66.74%.

Enterprise tenants access networks with committed bandwidth quotas shared among multiple departments and diverse services within each department. As a result, cloud vendors need to simultaneously fulfill two requirements, i.e., committed bandwidth guarantee and tenant-specified service bandwidth allocation. Hierarchical bandwidth allocation is a widely used technology that satisfies both requirements. In traditional schemes, each tenant's traffic is processed by a single node, potentially leading to single-node failures. Previous works have enhanced reliability by extending existing schemes to distributed systems with tenant-level bandwidth allocation, but fail to meet both requirements simultaneously. To bridge this gap, we propose S-DAL, which can achieve both requirements through servicelevel distributed hierarchical bandwidth allocation. We introduce an efficient fluid model-based algorithm for bandwidth allocation and employ a memory utilization based flow rate estimation mechanism to deliver accurate flow rate measurements. Additionally, we integrate a burst detection to mitigate excessive packet loss caused by burst traffic. Through testbeds and simulations, we demonstrate that S-DAL effectively ensures tenant-specified service bandwidth allocation while only reducing the shortfall in committed bandwidth to less than 0.23%.

Cross-domain object detection in remote sensing suffers from substantial domain gaps arising from differences in resolution, viewing geometry, and imaging modality across sensors and platforms. Existing unsupervised domain adaptive object detection (DAOD) methods typically align source and target features using the detector’s own target-domain representations. However, the extraction of these representations is constrained by the very domain discrepancies they aim to bridge, resulting in noisy and biased features that make alignment unstable. To address this limitation, we propose the Vision-Language model Supervised Domain Adaptor (VLSDA), a domain adaptation framework supervised by a frozen vision-language model (VLM). It leverages a frozen VLM image encoder as an additional and stable semantic domain to guide domain alignment. Our VLM-supervised Prototypical Alignment (VLPA) module stabilizes category-wise alignment through a tri-domain adversarial strategy that jointly aligns source-VLM, target-VLM, and source-target distributions. Complementing this, the Global Cross-domain Contrastive Alignment (GCCA) module enhances intra-class compactness and inter-class separability via supervised contrastive learning. Without requiring any fine-tuning of the VLM, our framework directly mitigates reliance on noisy target features and improves robustness to large distribution shifts. Extensive experiments on multiple cross-domain remote sensing benchmarks demonstrate consistent improvements over state-of-the-art methods, including 68.3% mAP50on xView→DOTA and 70.5% on HRRSD → SSDD. The code is available at https://github.com/JunhongLu0704/VLSDA.

Remote sensing images often contain objects with diverse orientations, making rotated object detectors particularly suitable for remote sensing object detection. Refined single-stage detectors, comprising two single-stage detectors and a feature alignment module, have achieved significant advancements. However, aligning features of large-scale rotated objects against complex backgrounds in remote sensing images remains challenging. To overcome this, we have enhanced the feature alignment module and proposed the Deformable Multi-scale Refined Rotation RetinaNet (DMR3Det). It begins by identifying rotational key points of objects against background and then employs an attention mechanism to integrate these pivotal features for refined detection, achieving global multi-scale feature extraction of rotated objects in remote sensing images. Experiments on DOTA1.0 demonstrate that DMR3Det outperforms recent single-stage detectors like Oriented RepPoints by over 1.7% in mean Average Precision (mAP).