With the proliferation of Internet of Things (IoT) applications, advanced time series forecasting techniques have become increasingly critical for managing and responding to complex temporal dynamics. However, traditional RNN-based methods have faced challenges in the Long-term Time Series Forecasting (LTSF) domain when dealing with excessively long look-back windows and forecast horizons. Consequently, the dominance in this domain has shifted towards Transformer, MLP, and CNN approaches. The substantial number of recurrent iterations are the fundamental reasons behind the limitations of RNNs in LTSF. To address these issues, we propose two novel strategies to reduce the number of iterations in RNNs for LTSF tasks: Segment-wise Iterations and Parallel Multi-step Forecasting (PMF). RNNs that combine these strategies, called SegRNN, significantly reduce the required recurrent iterations for LTSF, resulting in notable improvements in forecast accuracy and inference speed. Extensive experiments demonstrate that SegRNN not only outperforms state-of-the-art Transformer-based models but also reduces runtime and memory usage by more than 78%, making it highly suitable for resource-constrained IoT scenarios. These achievements provide strong evidence that RNNs continue to excel in LTSF tasks and encourage further exploration of this domain with more RNN-based approaches. The code is available at: https://github.com/lss-1138/SegRNN.

Accurately forecasting power consumption in data center servers requires addressing the temporal distribution shift caused by dynamic resource demands. However, existing methods rely on global normalization, which cannot capture short-term localized shift, leading to unsatisfactory performance when forecasting non-stationary time series. To address this challenge, we propose a novel bi-level optimization framework for forecasting non-stationary long-term power consumption, namedNoSPF. The framework employs hierarchical optimization to separately model the stationary and local non-stationary components of power consumption time series, offering a flexible, model-agnostic paradigm for time-series forecasting. Using Discrete Wavelet Transform (DWT) for multi-scale time–frequency analysis,NoSPFdecomposes the series into non-stationary components driven by short-term fluctuations and stationary components that capture long-term trends. Furthermore,NoSPFintegrates a lightweight Multi-Layer Perceptron (MLP) to predict the local non-stationary components, enhancing the framework’s forecasting accuracy by providing more precise approximations of the future power distribution. Extensive experiments on real-world server datasets demonstrate the superior performance and effectiveness ofNoSPF.

Nowadays, machine learning (ML)-based power prediction models for servers have shown remarkable performance, leveraging large volumes of labeled data for training. However, collecting extensive labeled power data from servers in cloud data centers incurs substantial costs. Additionally, varying resource demands across different workloads (e.g., CPU-intensive, memory-intensive, and I/O-intensive) lead to significant differences in power consumption behaviors, known as domain shift. Consequently, power data collected from one type of workload cannot effectively train power prediction models for other workloads, limiting the exploration of the collected power data. To tackle these challenges, we proposeTGCP, a cross-workload power prediction method based on multi-source transfer Gaussian process regression.TGCPtransfers knowledge from abundant power data across multiple source workloads to a target workload with limited power data. Furthermore, Continuous normalizing flows adjust the posterior prediction distribution of Gaussian process, making it locally non-Gaussian, enhancingTGCP's ability to handle real-world power data distribution. This method enhances prediction accuracy for the target workload while reducing the expense of acquiring power data for real cloud data centers. Experimental results on a realistic power consumption dataset demonstrate thatTGCPsurpasses four traditional ML methods and three transfer learning methods in cross-workload power prediction.

Harmful fine-tuning attack poses a serious threat to the online fine-tuning service. Vaccine, a recent alignment-stage defense, applies uniform perturbation to all layers of embedding to make the model robust to the simulated embedding drift. However, applying layer-wise uniform perturbation may lead to excess perturbations for some particular non-safety-critical layers, resulting in defense performance degradation and unnecessary memory consumption. To address this limitation, we propose a Targeted Vaccine (T-Vaccine), a memory-efficient safety alignment method that applies perturbation to only selected layers of the model. T-Vaccine follows two core steps: First, it uses the harmful gradient norm as a statistical metric to identify the safety-critical layers. Second, instead of applying uniform perturbation across all layers, T-Vaccine only applies perturbation to the safety-critical layers while keeping other layers frozen during training. Results show that T-Vaccine outperforms Vaccine in terms of both defense effectiveness and resource efficiency. Comparison with other defense baselines, e.g., RepNoise and TAR also demonstrate the superiority of T-Vaccine. Notably, T-Vaccine is the first defense that enables a fine-tuning-based alignment method for 7B pre-trained models trained on consumer GPUs with limited memory (e.g., RTX 4090).

As deep learning (DL) workloads scale in complexity and volume, ensuring predictable job queuing times has become a critical challenge for data centers. Existing scheduling solutions primarily focus on minimizing tardiness or job completion times (JCT), often neglecting the need for deterministic queuing, particularly in dynamic and preemptive environments. This paper introducesKairos, a preemption-based scheduling framework enhanced by user collaboration to address these gaps.Kairoscombines adivide-and-conquerstrategy—segmenting jobs into sequential units with adaptive priorities—and a user-collaborative mechanism for better duration estimation. By leveraging real-time feedback from resource contention and queuing delays,Kairosminimizes a novel metric, theQueue inStability Index(QSI), achieving significant improvements in queuing predictability while maintaining competitive JCT. Experimental results demonstrate thatKairosreduces QSI by over 99.8% compared to state-of-the-art deadline-aware baselines, offering robust performance for diverse DL workloads.

The stable periodic patterns present in time series data serve as the foundation for conducting long-horizon forecasts. In this paper, we pioneer the exploration of explicitly modeling this periodicity to enhance the performance of models in long-term time series forecasting (LTSF) tasks. Specifically, we introduce the Residual Cycle Forecasting (RCF) technique, which utilizes learnable recurrent cycles to model the inherent periodic patterns within sequences, and then performs predictions on the residual components of the modeled cycles. Combining RCF with a Linear layer or a shallow MLP forms the simple yet powerful method proposed in this paper, called CycleNet. CycleNet achieves state-of-the-art prediction accuracy in multiple domains including electricity, weather, and energy, while offering significant efficiency advantages by reducing over 90% of the required parameter quantity. Furthermore, as a novel plug-and-play technique, the RCF can also significantly improve the prediction accuracy of existing models, including PatchTST and iTransformer. The source code is available at: https://github.com/ACAT-SCUT/CycleNet.

The cyber-physical cloud systems (CPCSs) release powerful capability in provisioning the complicated industrial services. Due to the advances of machine learning (ML) in attack detection, a wide range of ML applications are involved in industrial CPCSs. However, how to ensure the implementation efficiency of these applications, and meanwhile avoid the privacy disclosure of the datasets due to data acquisition by different operators, remain challenging for the design of the CPCSs. To fill this gap, in this article a privacy-aware deployment method (PDM), named PDM, is devised for hosting the ML applications in the industrial CPCSs. In PDM, the ML applications are partitioned as multiple computing tasks with certain execution order, like workflows. Specifically, the deployment problem is formulated as a multiobjective problem for improving the implementation performance and resource utility. Then, the most balanced and optimal strategy is selected by leveraging an improved differential evolution technique. Finally, through comprehensive experiments and comparison analysis, PDM is fully evaluated.

Nowadays, Edge computing is being introduced as a powerful paradigm to collaborate with the cloud to provide sufficient computing resources for IoT to implement intelligent analysis and data mining. Generally, due to the edge nodes are closer to mobile users, the access latency and the cost of using cloud services are effectively reduced. However, the implementation of cross-layer resource scheduling between edge nodes and servers deployed in the cloud to meet service requirements (i.e., shortest completion time, maximum resource utilization, lower energy consumption, etc.) still faces great challenges. To address this challenge, a cross-layer resource scheduling method, named CRSM, for the IoT applications is proposed in this paper. Technically, the Pareto archived evolution strategy (PAES) is employed to optimize the time cost of IoT applications, resource utilization and energy consumption of edge node. Then, the technique for order preference by similarity to ideal solution (TOPSIS) and the multiple criteria decision making (MCDM) are leveraged to acquired the optimal cross-layer resource scheduling strategy. Finally, the comprehensive analysis of CRSM is introduced in detail.

Cloud computing is a formidable paradigm to provide resources for handling the services from Industrial Internet of Things (IIoT), such as meteorological industry. Generally, the meteorological services, with complex interdependent logics, are modeled as workflows. When any of the computing nodes for hosting the meteorological workflows fail, all sorts of consequences (e.g., data loss, makespan enlargement, performance degradation, etc.) could arise. Thus recovering the failed tasks as well as optimizing the makespan and the load balance of the computing nodes is still a critical challenge. To address this challenge, a dynamic resource provisioning method (DRPM) with fault tolerance for the data-intensive meteorological workflows is proposed in this article. Technically, the Virtual Layer 2 (VL2) network topology is exploited to build meteorological cloud infrastructure. Then, the nondominated sorting genetic algorithm II (NSGA-II) is employed to minimize the makespan and improve the load balance. Finally, comprehensive experimental analysis of DRPM are proceeded.