Concurrency and Computation-Practice & Experience最新文献

For cross-city traffic prediction, the significant heterogeneity of traffic data across cities and the requirement for privacy protection make it challenging for conventional centralized spatiotemporal graph modeling techniques to balance predictive performance and data security. Therefore, this paper proposes AT-SPNet, a personalized federated spatiotemporal modeling approach specifically designed for cross-city traffic prediction. This method decouples the spatiotemporal modeling paths through the construction of a shared temporal branch and a hidden local spatial branch, thereby mitigating the heterogeneity of cross-city traffic data while preserving privacy. In the temporal branch, Gated Recurrent Units and a multi-head attention mechanism are incorporated to capture temporal dependencies, and a Squeeze-and-Excitation module is employed to enhance the extraction of informative features. In the spatial branch, a Spatial Attention Fusion module based on a triple-attention mechanism is designed to capture spatial features from multiple spatial perspectives, combined with static graph convolution and dynamic graph attention to construct a dual-modal information fusion path. Furthermore, to alleviate the adverse effects of cross-city data heterogeneity in federated training, a personalized federated learning strategy is introduced, which enables differentiated fusion of client spatial features without sharing raw data. Experiments on four real-world traffic datasets demonstrate that AT-SPNet outperforms existing methods in both prediction accuracy and cross-city generalization, validating the effectiveness and practical applicability of the proposed approach for cross-city traffic prediction.

Remote procedure call (RPC) technology has become a cornerstone of modern cloud computing, enabling efficient and seamless communication between distributed services. In cloud infrastructures, where scalability, interoperability, and performance are critical, RPC frameworks play a key role in abstracting the complexities of network communication. Despite their ubiquity, relatively little up-to-date empirical research exists on the comparative performance of RPC frameworks across cloud environments. And, to our best knowledge, there is no comparative study in which different cloud platforms and RPC frameworks were directly compared. This paper addresses that gap by presenting the results of a series of experiments evaluating the performance and scalability of four major RPC frameworks—ONC RPC, gRPC, Web-RPC, and JSON-RPC—across the three most widely used cloud platforms: AWS, Azure, and Google Cloud. The experiments were based on a test suite comprising four distinct RPC call types with varying argument sizes and complexities. Each configuration was executed multiple times with client loads ranging from 1 to 8, with 300,000 runs per test. Total call latency was used as the primary performance measure and analyzed statistically. The results reveal a nuanced picture: while ONC RPC consistently delivers the best performance, no other framework or platform emerges as a clear overall leader. Across all cloud environments and workloads, ONC RPC consistently outperformed the other frameworks. It proved to be at least 20% faster than its nearest competitor and, in some cases, up to four times faster than the second best. By contrast, in short-argument tests, gRPC performed unexpectedly poorly; it often ranked close to or below the text-based frameworks. Particularly surprising was its poor performance under high load in Google Cloud—the platform where it could be expected to perform best. However, gRPC ranked second in the large-argument tests, while text-based frameworks show relatively poor performance as the argument size increases. We discuss and explain these findings in detail and provide guidelines for selecting the most suitable RPC technology for different use cases.

Public auditing is a method used to verify the integrity of data stored in the cloud without requiring access to the actual data. However, the advancement of quantum computers poses significant security threats to existing public auditing schemes, as these schemes are based on conventional cryptography hard problems, which are vulnerable to quantum attacks. To address this, NIST has launched the development of post-quantum cryptographic primitives and protocols. Among the various approaches, lattice-based cryptography (LBC) is considered one of the most promising candidates due to its strong security guarantees and inherent resistance to quantum attacks. Leveraging LBC, several researchers have proposed lattice-based public auditing (LBPA) schemes for cloud storage security based on lattice hardness assumptions. This paper provides a comprehensive survey of existing LBPA schemes for cloud storage, presenting a detailed taxonomy and analyzing their similarities, differences, and performance. Additionally, it highlights key challenges and outlines future research directions for designing efficient and secure public auditing schemes in the post-quantum era.

With the exponential growth of graph-structured data, single-machine efficient analysis has become increasingly impractical, making high-performance distributed graph computing systems indispensable. The efficacy of these systems hinges critically on high-quality graph partitioning. The edges of many graphs stemmed from real applications are uncertain, but many existing graph partitioning algorithms are only for deterministic graphs without considering uncertainty. This paper presents a novel partitioning algorithm, PAUG (Partitioning Algorithm for Uncertain Graphs), tailored for uncertain graphs. First, it formalizes the partitioning problem as an optimization task to minimize the cut-edge ratio while balancing load. Second, it introduces probabilistic similarity to quantify vertex relationships under uncertainty. Finally, it details the PAUG algorithm which consists of initial partition phase and score-function-guided refinement strategy. Experimental results shows that PAUG achieves an average 23.2% reduction in cut-edge ratio and a 26.2% improvement in load balance over state-of-the-art algorithms.

Cloud computing offers on-demand access to computing resources; however, minimizing response time while ensuring compliance with Service Level Agreements (SLAs) remains a critical challenge. The proposed CLOUD SMART framework aims to intelligently minimize response time, process delays, and propagation latency in cloud environments through adaptive, SLA-aware dynamic scheduling. This study evaluates the model using the Cloud Workload Dataset for Scheduling Analysis, available on Kaggle. The dataset undergoes preprocessing, including median imputation for missing numerical values, and the derivation of key features such as response time, deadlines, and priority tiers for accurate workload profiling. Workload characterization through statistical profiling and clustering reveals patterns, arrival rates, and task categories that guide scheduling strategy selection. Baseline performance is established using discrete event simulation of standard policies such as FCFS, SJF/Min-Min, Max-Min, and EDF. A novel Predictive Deadline-Aware Hybrid Scheduling (PDHS) approach is integrated to predict completion times and dynamically switch scheduling strategies based on urgency. An execution and closed-loop feedback mechanism enables real-time adaptation. Experimental results show that CLOUD SMART significantly reduces response time, improves SLA compliance, and enhances resource utilization compared to static scheduling baselines. The PDHS model achieves an average response time of 6.72 s, significantly lower than all baselines. Average waiting time is reduced to 2.87 s, and Makespan improves to 138.4 s. SLA compliance reaches 97%, with a deadline miss ratio of only 3%. System throughput is enhanced to 38.5 tasks per second, and resource utilization climbs to 92%. Prediction accuracy excels with an MAE of 0.94 s and RMSE of 1.26 s.

Intent-based networking (IBN) is an advanced approach that combines artificial intelligence (AI) and advanced machine learning (ML) technologies with automation for advanced network management by adapting the network's functions to an organization's business objectives. The concept of IBN is transforming autonomous network management with regard to cybersecurity. It improves the detection, response, and prevention of threats by encoding security policies into automated network actions and configurations. IBN enables adaptive, proactive, and enduring protection in sophisticated and evolving cybersecurity environments, thereby reinforcing cybersecurity resilience. This work proposes an integrated approach that combines IBN with ML and reinforcement learning (RL) to effectively defend against TCP SYN-based DDoS attacks. The ML model proposed in this research outputs a detection accuracy of 99.86%. Additionally, the RL-based response mitigation strategy improves response time by 43% in comparison to traditional reactive security methods. In the architecture IBNS framework, it actively updates high-level security intents into adaptive policies for near real time response to mitigate threats achieving less than 0.0008 false positive rate. Besides, the system is able to remain stable within the network while automating the response to the threats and enduring versatile attacks. This automated approach allows security to be aligned with operational objectives enabling proactive, adaptive, and multi-dimensional security. These findings mark the advancement that intelligent autonomous systems can provide towards cybersecurity in future network infrastructures, showing benefits of enhancing resilience through self-learning and intent-driven mechanisms.

Memory models define the order in which accesses to shared memory in a concurrent system may be observed to occur. Such models are a necessity since program order is not a reliable indicator of execution order, due to microarchitectural features or compiler transformations. Concurrent programming, already a challenging task, is thus made even harder when weak memory effects must be addressed. A rigorous specification of weak memory models is therefore essential to make this problem tractable for developers of safety- and security-critical, low-level software. In this paper we survey the field of formalisations of weak memory models, including their specification, their effects on execution, and tools and inference systems for reasoning about code. To assist the discussion we also provide an introduction to two styles of formal representation found commonly in the literature (using a much simplified version of Intel's x86 as the example): a step-by-step construction of traces of the system (operational semantics); and with respect to relations between memory events (axiomatic semantics). The survey covers some long-standing hardware features that lead to observable weak behaviours, a description of historical developments in practice and in theory, an overview of computability and complexity results, and outlines current and future directions in the field.

We propose an incremental similarity-based label propagation algorithm (DLPA-S) for detecting dynamic community structures. As the network evolves, the method efficiently updates the communities over time via local label updates driven by changes in network topology—including edge and vertex additions or removals—and vertex similarity. This incremental approach significantly reduces computational cost while preserving accuracy in capturing community evolution. We evaluate DLPA-S using a comprehensive set of quality metrics that assess both the structural properties of the network and the agreement between detected communities and ground-truth partitions. Experiments are conducted on synthetic and real-world dynamic networks, varying key graph characteristics such as the number of vertices and the average degree, as well as across diverse community scenarios. The results show that DLPA-S consistently achieves stable and high-performing results, maintains high NMI and F1 scores, ensures strong internal connectivity, clear community separability, and avoids disconnected communities, while remaining computationally efficient.