When processing large-scale co-located inference workloads in heterogeneous GPU clusters, existing cluster scheduling mechanisms often increase the job makespan due to neglecting the mutual performance interference and resource contention for co-located tasks. This deficiency is dramatically amplified in typical neural networks based deep learning workloads because these workloads are highly sensitive to hardware configurations like GPU memory capacity, bandwidth and GPU core frequency. Therefore, a hardware-performance-aware scheduler capable of adaptively and dynamically dispatching tasks according to GPU hardware characteristics is crucial for reducing the overall completion time of job queues. To address this issue, we propose Hardware-Aware Scheduling for Inference Tasks in Heterogeneous GPU Clusters (HASE), a novel scheduling framework that dynamically adapts to GPU hardware characteristics for real-time optimal task placement.HASEconsists of two core components: a kernel-level latency prediction model and a hybrid scheduling strategy. Unlike conventional approaches that rely on coarse-grained model-level features, our predictor decomposes inference models into fine-grained computational kernels using ONNX Runtime graph optimization, and predicts individual kernel execution times under varying GPU load conditions. By integrating static hardware specifications, dynamic microbenchmarks, and real-time DCGM profiling metrics, the predictor captures both operator-level heterogeneity and background load interference. The hybrid scheduling strategy combines a two-stage greedy search for task placement with a resource reservation and backfilling mechanism to balance immediate optimization and long-term fairness. Experimental results on CNN-based inference workloads (YOLO, ResNet, VGG, DenseNet, and MobileNet series) demonstrate thatHASEachieves a kernel-level prediction accuracy of 8.2% MAPE and model-level accuracy of$R^{2}$= 0.91. Moreover,HASEreduces 51% total job makespan compared to traditional round-robin scheduling, and maintains per-task scheduling decision time under one second in clusters with up to hundreds of GPUs.

Traditional log anomaly detection systems are centralized, which poses the risk of privacy leakage during data transmission. Previous research mainly focuses on single-domain logs, requiring domain-specific models and retraining, which limits flexibility and scalability. In this paper, we propose a unified federated cross-domain log anomaly detection approach, FLUXLog, which is based on MoE (Mixture of Experts) to handle heterogeneous log data. Based on our insights, we establish a two phase training process: pre-training the gating network to assign expert weights based on data distribution, followed by expert driven top-down feature fusion. The following training of the gating network is based on fine-tuning the adapters, providing the necessary flexibility for the model to adapt across domains while maintaining expert specialization. This training paradigm enables a Hybrid Specialization Strategy, fostering both domain-specific expertise and cross-domain generalization. The Cross Gated Experts Module (CGEM) then fuses expert weights and dual channel outputs. Experiments on public datasets demonstrate that our model outperforms baseline models in handling unified cross-domain log data.

Microservice architecture, characterized by loose coupling, elastic scalability, and independent deployment, has become the de facto standard for cloud-native services. However, in production systems, the complex interaction topologies among microservices can easily trigger cascading failures, which severely compromise system reliability and operational stability. This paper proposes an anomaly detection method for microser vices that integrates adaptive wavelet convolutional networks and Transformer architecture. Specifically, the proposed method leverages a wavelet convolutional network to preprocess time series data, while a weighted inverse transformer combined with a frequency-band attention mechanism is utilized to capture multi scale time–frequency features of different microservices. Moreover, a dual-branch GRU-Transformer structure is then used to model both long- and short-term dependencies as well as cross metric correlations among services. Therefore, anomaly scores can be derived through the joint optimization of prediction and reconstruction models. To overcome the challenge of threshold determination, this paper also proposes a two-stage gradient based optimization algorithm that adaptively converges the threshold range using finite differential methods. Experimental results show that the proposed method is effective on datasets such as SMD (Server Machine Dataset), ASD (Application Server Dataset) and the NASA Anomaly Detection Dataset SMAP, where its average F1 score on SMD exceeds that of existing approaches by 11.0%, while on the SMAP dataset, an accuracy of 96.40% is achieved. These results confirm that the proposed approach can effectively detect anomalies in industrial level microservices, thereby enhancing the stability and reliability of industrial service systems.

Edge computing and Internet of Things (IoT) devices have been widely deployed in smart city applications due to the rapid promotion and implementation of 5G communication technology. Limited by power supply and hardware computing capability, applications on edge servers are mostly deployed and run in the form of container microservices to improve resource utilization. Therefore, identifying the power consumption at the container granularity is of great importance for the load scheduling and service quality assurance of edge servers. In this article, a gradient boosting piecewise linear regression tree (GBRT-PL)-based container power prediction method is proposed. Performance metrics with a strong correlation between container and server power consumption are selected for power consumption modeling. To effectively suit the nonlinear relationship between container performance metrics and server power consumption, the integrated predictive capabilities of multiple regression trees (RTs) and the segmented linear model of single RT leaf nodes are applied. The data from the experiments prove that the GBRT-PL model predicts power consumption more accurately than other models for single and multiple container groups. The highest average relative error rate in the four multicontainer group tests is 6.72%, whereas the highest relative error rate in the 90% quantile is 11.66%. In addition, it can accurately predict the majority of power consumption peaks, which contributes to the precise detection of power consumption anomalies.

Workload characteristics are vital for both data center operation and job scheduling in co-located data centers, where online services and batch jobs are deployed on the same production cluster. In this article, a comprehensive analysis is conducted on Alibaba's cluster-trace-v2018 of a production cluster of 4034 machines. The findings and insights are the following: (1) The workload on the production cluster poses a daily cyclical fluctuation, in terms of CPU and disk I/O utilization, and the memory system has become the performance bottleneck of a co-located cluster. (2) Batch jobs including their tasks and derived instances can be approximated as Zipf distribution. However, for all batch jobs with directed acyclic graph dependency, they suffer from co-location with online services since the online services are highly prioritized. (3) The resource usages of containers have similar cyclical fluctuation consistent with the whole cluster, while their memory usages remain approximately constant. (4) The number of batch jobs co-located with online services is dependent on the mispredictions per kilo instructions of online services. In order to guarantee the QoS of online services, when the MPKI of online services rises, the number of batch jobs to be co-located on the same machine should decrease.

In this article, we provide a concise but systematic review on blockchain-enabled cyber-physical systems (CPS). We dissect various blockchain-enabled CPS as reported in the literature in terms of their operations and the features of blockchain that have been used. We identify key common CPS operations that can be enabled by blockchain, and classify them in terms of their time sensitivity and throughput requirements. We also elaborate and classify features of blockchain in terms of different levels of benefits to CPS, including security, privacy, immutability, fault tolerance, interoperability, data provenance, atomicity, automation, data/service sharing, and trust. Finally, we point out two primary open research issues for developing blockchain-enabled CPS, namely, excessive delay in reaching consensus and limited throughput, and outline future research directions.

The traditional centralized network architecture can lead to a bandwidth bottleneck in the core network. In contrast, in the information‐centric network, decentralized in‐network caching can alleviate the traffic flow pressure from the network center to the edge. In this paper, a popularity‐aware in‐network caching policy, namely, Pop, is proposed to achieve an optimal caching of network contents in the resource‐constrained edge networks. Specifically, Pop senses content popularity and distributes content caching without adding additional hardware and traffic overhead. We conduct extensive performance evaluation experiments by using ndnSIM. The experiments showed that the Pop policy achieves 54.39% cloud service hit reduction ratio and 22.76% user request average hop reduction ratio and outperforms other policies including Leave Copy Everywhere, Leave Copy Down, Probabilistic Caching, and Random choice caching. In addition, we proposed an ideal caching policy (Ideal) as a baseline whose popularity is known in advance; the gap of Pop and Ideal in cloud service hit reduction ratio is 4.36%, and the gap in user request average hop reduction ratio is only 1.47%. More simulation results further show the accuracy of Pop in perceiving popularity of contents, and Pop has good robustness in different request scenarios.

With the development of big data, big data processing systems, such as Hadoop and Spark, are widely used to handle large-scale data. To avoid the complexity and expensiveness of building a self-owned big data processing system, cloud providers tend to deploy big data processing tools as cloud services. Typical examples include Amazon EMR, Azure HDInsight and AliCloud E-MapReduce. However, how to build a cost-efficient system and scale the system is still challenging. In this paper, we have conducted a case study on AliCloud E-MapReduce, and analyzed the system performance upon local and remote file systems. We compared the scalability of Hadoop and Spark by using scaleout and scale-up strategies respectively. Based on the analysis results, we derive several observations and implications, which will contribute to guide the performance optimization.

Servers have different energy efficiency and energy proportionality (EP) due to their hardware configuration (i.e., CPU generation and memory installation) and workload. However, current virtual machine (VM) scheduling in virtualized environments will saturate servers without considering their energy efficiency and EP differences. This article will discuss EASE, the energy efficiency and proportionality aware VM scheduling approach. EASE first executes customized computing intensive, memory intensive, and hybrid benchmarks to calculate a server's energy efficiency and EP. Then it schedules VMs to servers to keep them working at their peak energy efficiency point (or optimal working range). This step improves the overall energy efficiency of the cluster and the data center. For performance guarantee, EASE migrates VMs from servers under highly contending conditions. The experimental results on real clusters show that power consumption can be saved 37.07% ~ 49.98% in the homogeneous cluster. The average completion time of the computing intensive VMs increases only 0.31 % ~ 8.49%. In the heterogeneous nodes, the power consumption of the computing intensive VMs can be reduced by 44.22 %. The job completion time can be saved by 53.80%.

The authors propose to use formatting templates and implicit formatting semantics information for automatic metadata identification and segmentation. The pure texts and their corresponding formatting information including line height, font type, and font size, are recognized in parallel to guide metadata identification. The authors use implicit formatting semantics, such as the change of formatting, formatting templates and implications, explicit formatting layouts, as well as predefined frequently occurred keywords database to increase the extraction accuracy. Unlike other OCR-based approaches, the authors use open source PDFBox package as the basic preprocessing tool to get pure texts and formatting values of the document contents. On top of PDFBox they built their own pipeline program, namely, PAXAT, to implement their approaches for metadata extraction. 10177 papers from arXiv, ACM, ACL and other publicly accessed and institution-subscribed sources are tested. The overall extraction accuracy of title, authors, affiliations, author-affiliation matching are 0.9798, 0.9425, 0.9298, and 0.9109, respectively.

The huge energy consumption in data centers produces not only high electricity bill but also tremendous carbon footprints. Although today's servers and data centers of leading internet companies are more energy efficient than ever before, the fluctuations in external workload and internal resource utilization calls for energy proportional computing. Insight into server energy proportionality can help improve workload placement while also reducing energy consumption. In this paper, we investigate all 477 valid published results of SPECpower_ssj benchmark from 2007 to 2016Q3 and reorganize them by hardware availability year for more accurate analysis on production servers. Through comprehensive analysis we find that: (1) The specious stagnation of energy proportionality in recent years is mainly caused by the adoption of processors of specific microarchitecture and is not the indicative trend of energy proportionality improvement. (2) Microarchitecture evolution has more influence on energy efficiency improvement than energy proportionality. (3) Today's servers' peak energy efficiencies are shifting from 100% resource utilization to 80% or 70% utilization and server energy proportionality improves with such shifting. We then conduct extensive experiments on 4 rack servers to investigate the energy efficiency variations under different hardware configurations, including memory per core installation and processor frequency scaling. Our experiments show that hardware configuration has significant impact on server's energy efficiency. Our findings presented in this paper provide useful insights and guidance to system designers, as well as data center operators for energy proportionality aware workload placement and energy savings.

Summary Parallel semi‐implicit method for pressure‐linked equations(SIMPLE) algorithm is used to solve the 3‐D incompressible pipe flow problem. In this paper, we proposed a novel parallel SIMPLE algorithm that uses the alternate tiling technique. Firstly, a parallel SIMPLE algorithm based on domain decomposition method was established, and the implementation of domain partition and data exchange was presented. Then, we presented serial finite difference stencil algorithm based on alternate tiling. Furthermore, an iteration space parallel two‐way finite difference stencil algorithm based on alternate tiling was proposed, introducing the sequence of iterative space tiles as the sequence of execution and using time skewing technique to partition the iteration space, thus to improve the data locality of algorithm. The cache misses and the cost of communication and synchronization are reduced by reordering the tiles of iteration space. Finally, the effectiveness of the two parallel SIMPLE algorithms were compared. The results showed that the parallel SIMPLE algorithm that uses the two‐way finite difference stencil algorithm based on alternate tiling has good data locality, performance, and scalability in the Deepcomp7000 cluster computing environment. Copyright © 2013 John Wiley & Sons, Ltd.

SIMPLE (Semi-Implicit Method for Pressure-Linked Equations) algorithm is important in the simulation of steady flows. As the traditional 3-D SIMPLE algorithm is time-consuming, we propose a parallel SIMPLE algorithm based on a novel tiling strategy -- alternate tiling, through replacing the original linear system and reordering the iteration space tiles. The novelty of our parallel algorithm lies in the introduction of the sequence of iteration space tiles as the sequence of execution, the time skewing technique to partition the iteration space, update operations of the grids from two directions alternately, and the improvement of the data locality. The effectiveness of the parallel algorithm and serial model of finite difference stencil algorithm are validated. Numerical experiments on distributed clusters show that the cache misses and the cost of communication and synchronization are reduced by reordering the tiles of iteration space, and the parallel SIMPLE algorithm based on alternate tiling has a good data locality and parallel efficiency in the three-dimensional incompressible pipe flow project.

Due to various differences in hardware architectures of devices in ubiquitous computing systems, portability and platform-independency become the main challenge for graphics programming in system design. In this paper, we propose an adaptive user interface programming toolkit for system design in ubiquitous computing environment. The toolkit leverages an existing system software infrastructure, making the application programming straightforward and platform independent. This proposed toolkit can be divided into two parts: the first part consists of open source cross-platform graphics libraries which are encapsulated into the platform dependent part of backend library for interacting with specific system. While another one, called core library, is responsible for the functions of control logics, graphics drawing and backend management. To demonstrate the practical use of this toolkit and its portability, a case study is provided for demonstration. The test results on three different embedded systems show its good adaptability on multi-platforms.

With the scale of computing system increases, power consumption has become the major challenge to system performance, reliability and IT management costs. Specifically, system performance and reliability, described by various Quality of Service(QoS) metrics, cannot be guaranteed if the objective is to minimize the total power consumption solely, despite of the violations of QoS. Various methods have been developed to control power consumption to avoid system failures and thermal emergencies through coarse-grained designs. However, the existing methods can be improved and more power can be saved if fine-grained job level adaptation is integrated into them. In this paper a feedback control based power aware job scheduling algorithm is proposed to minimize power consumption in computing system and to provide QoS guarantees. In the proposed algorithm, jobs are scheduled according to the realtime and historical power consumption as well as the QoS requirements. Simulations and experiments on real multi core computing system show that the power potential of the system can be deeply explored while still providing QoS guarantees and the performance degradation is acceptable. The experiment results also show that fine-grained job-level power aware scheduling can achieve better power/performance balancing between multiple processors or cores than coarse-grained methods.

Wireless sensor networks (WSNs) are emerging as essential and popular ways of providing pervasive computing environments for various applications. In all these environments energy constraint is the most critical problem that must be considered. Clustering is introduced to WSNs because of its network scalability, energy-saving attributes and network topology stabilities. However, there also exist some disadvantages associated with individual clustering scheme, such as additional overheads during cluster-head (CH) selection, assignment and cluster construction process. In this paper, we discuss and compare several aspects and characteristics of some widely explored clustering algorithms in WSNs, e.g. clustering timings, attributes, metrics, advantages and disadvantages of corresponding clustering algorithms. This paper also presents a discussion on the future research topics and the challenges of clustering in WSNs.

Grids provide infrastructures and solutions to solve large scale cooperative problems such as large scale distributed virtual environment simulation, multi-institutional scientific computing and data analysis, etc. In this paper, the key techniques for Grid computing based large scale Distributed Cooperative Virtual Environment Simulation (GDCVES) are discussed and a hierarchical architecture of GDCVES is proposed. The solutions of GDCVES, such as resource management, massive data management, security aware task scheduling, and fault-tolerance are also discussed. To evaluate the feasibility and scalability of GDCVES, a prototype was implemented and the simulation workflow framework is also analyzed.

In this paper, a new genetic algorithm for solving multi-constrained optimization problems based on KS function is proposed. Firstly, utilizing the agglomeration features of KS function, all constraints of optimization problems are agglomerated to only one constraint. Then, we use genetic algorithm to solve the optimization problem after the compression of constraints. Finally, the simulation results on benchmark functions show the efficiency of our algorithm.