Circuit cutting decomposes a large quantum circuit into smaller subcircuits that are executed independently; the original circuit’s expectation values are then recovered by classically combining the measured subcircuit outcomes. While prior work characterises cutting overhead in terms of subcircuit counts and sampling complexity, its end-to-end impact on iterative, estimator-driven training pipelines remains insufficiently measured from a systems perspective. We propose DistributedEstimator, a cut-aware estimator execution pipeline that treats circuit cutting as a staged distributed workload. Each estimator query is instrumented across four phases: partitioning, subexperiment generation, parallel execution, and classical reconstruction. Using logged runtime traces and learning outcomes on two binary classification workloads (Iris and MNIST), we quantify cutting overheads, scaling limits, and sensitivity to injected stragglers, and evaluate whether accuracy and robustness are preserved under matched training budgets. Our measurements reveal that reconstruction constitutes a dominant fraction of per-query time—reaching a median of 53% and a 95th percentile of 58% at three cuts—thereby bounding achievable speed-up under increased parallelism. Despite these overheads, test accuracy is fully preserved on Iris and maintained without systematic degradation on MNIST across all evaluated cut configurations. Robustness under Gaussian noise and FGSM perturbations is similarly preserved, with several cut configurations exhibiting comparable or improved robustness relative to the uncut baseline. The exponential growth of subexperiment counts with each additional cut (O(9c) for CNOT-based decomposition) represents a fundamental computational barrier that limits practical experimentation to small qubit counts with current methods. These results establish that practical scaling of circuit cutting for learning workloads requires reducing and overlapping reconstruction, designing scheduling policies for barrier-dominated critical paths, and developing computationally efficient reconstruction strategies for larger qubit counts.

With the rapid growth of artificial intelligence and large-scale model computing, the demand for GPUs in datacenters continues to increase, especially for large-scale training and inference tasks. Heterogeneous multi-GPU systems, which integrate GPUs with varying types and computational capabilities, have become critical computing resources. This leads to two main challenges. First, due to the differences in GPU performance and power consumption, task scheduling involves a complex multi-objective optimization to balance energy efficiency and performance. More importantly, the lack of coordinated mechanisms for multi-task sharing and energy-efficient resource management across heterogeneous GPUs can result in GPU overload or underutilization, leading to wasted resources and potential system risks. To address these challenges, we propose HESHARE, an energy-aware and efficient heterogeneous GPU framework for datacenters. First, we design an energy-aware task scheduling strategy that optimizes task allocation across different GPUs to achieve a balance between energy consumption and performance. Second, we introduce a GPU sharing optimization mechanism that adaptively configures MPS and DVFS settings for each GPU, enhancing resource utilization, reducing overall energy consumption, and ensuring task performance. Compared to the state-of-the-art framework, we reduce average energy costs by 26% and improve job completion time by 31%, achieving a balance between energy efficiency and performance.

In stream computing systems, fault tolerance and recovery efficiency during task execution are core elements for ensuring system performance. However, existing fault tolerance strategies often overemphasise global fault tolerance performance, leading to significant increases in resource overhead and failing to achieve an effective balance between resource costs and reliability. To address these issues, we propose a dynamic adaptive fault-tolerant strategy named Da-Stream. This paper addresses the following aspects: (1) The high resource costs associated with running both primary and backup copies simultaneously are analysed, and the impact of factors such as operator type, changes in data stream size, and the strength of upstream/downstream dependencies on the fault tolerance requirement level of operators is verified. (2) The stream computing resource model and operator fault tolerance requirement level model are established to evaluate node CPU and memory resource utilisation, satisfy the resource constraints of the fault tolerance strategy, and adaptively assess the fault tolerance requirement levels of different operators. (3) Da-Stream classifies strategies based on factors such as operator type, resource requirements, and dependencies, and dynamically adjusts the fault tolerance strategy at runtime by combining resource usage, fault prediction, and historical scores. (4) Experimental results show that compared with state-of-the-art methods, Da-Stream reduces fault recovery time by 24.3

The rapid proliferation of IoT applications has intensified the demand for efficient and secure service placement in Fog computing. However, heterogeneous resources, dynamic workloads, and diverse security requirements make optimal service placement highly challenging. Most solutions focus primarily on performance metrics while overlooking the security implications of deployment decisions. This paper proposes a Security and Performance-Aware Distributed Deep Reinforcement Learning (SPA-DDRL) framework for joint optimization of service response time and security compliance in Fog computing. The problem is formulated as a weighted multi-objective optimization task, minimizing latency while maximizing a security score derived from the security capabilities of Fog nodes. The security score features a new three-tier hierarchy, where configuration-level checks verify proper settings, capability-level assessments evaluate the resource security features, and control-level evaluations enforce stringent policies, thereby ensuring compliant solutions that align with performance objectives. SPA-DDRL adopts a distributed broker-learner architecture where multiple brokers perform autonomous service-placement decisions and a centralized learner coordinates global policy optimization through shared prioritized experiences. It integrates three key improvements, including Long Short-Term Memory networks, Prioritized Experience Replay, and off-policy correction mechanisms to improve the agent's performance. Experiments based on real IoT workloads show that SPA-DDRL significantly improves both service response time and placement security compared to current approaches, achieving a 16.3

Computation offloading at lower time and lower energy consumption is crucial for resource-constrained mobile devices. This paper proposes an offloading decision-making model using federated learning. Based on the device configuration, task type, and input, the proposed decision-making model predicts whether the task is computationally intensive or not. If the predicted result is computationally intensive, then based on the network parameters the proposed decision-making model predicts whether to offload or locally execute the task. The experimental results show that the proposed method achieves above 90% prediction accuracy in offloading decision-making, and reduces the response time and energy consumption of the user device by similar to 11-31 %. A secure partial computation offloading method for federated learning is also proposed to deal with the Straggler effect of federated learning. The results present that the proposed partial computation offloading method for federated learning has achieved a prediction accuracy of above 98% for the global model.

Matrix multiplication is a fundamental kernel in large-scale artificial intelligence and scientific computing, but its performance on conventional electronic accelerators is increasingly constrained by memory bandwidth and energy efficiency. Photonic computing offers a promising alternative due to its ultra-high bandwidth, massive parallelism, and low power dissipation. However, most existing photonic systems are limited to low-precision computation because of analog optical modulation constraints and noise accumulation, which restricts their applicability in precision-critical workloads. To address this limitation, we propose LightMat-HP, a hybrid photonic-electronic computing system that enables end-to-end acceleration of general matrix multiplication with configurable computational precision. LightMat-HP adopts block floating-point (BFP) arithmetic to reduce computational complexity while enabling flexible precision-performance tradeoffs. To overcome the precision limitations of photonic devices, we propose a slicing-based photonic multiplication scheme that exploits the high accuracy of low bit-width photonic multiplication in combination with digital accumulation to achieve high-precision mantissa multiplication. A tile-based matrix multiplication dataflow is further designed to support matrices of arbitrary sizes. We experimentally validate LightMat-HP on a photonic computing prototype and evaluate its performance through large-scale simulations. The results demonstrate that LightMat-HP outperforms FPGA, GPU, and a state-of-the-art photonic accelerator across throughput, latency, and energy efficiency, particularly for small- and medium-sized matrix multiplications, owing to its highly parallel photonic architecture, efficient data movement, and slice-based BFP arithmetic.

Applications are moving away from monolithic designs to microservice and serverless architectures, where fleets of lightweight and independently deployable components run on public clouds. Autoscaling serves as the primary control mechanism for balancing resource utilization and quality of service, yet existing policies are either opaque learned models that require substantial per-deployment training or brittle hand-tuned rules that fail to generalize. We investigate whether large language models can act as universal few-shot resource allocators that adapt across rapidly evolving microservice deployments. We propose ORACL, Optimized Reasoning for Autoscaling via Chain of Thought with LLMs for Microservices, a framework that leverages prior knowledge and chain-of-thought reasoning to diagnose performance regressions and recommend resource allocations. ORACL transforms runtime telemetry, including pods, replicas, CPU and memory usage, latency, service-level objectives, and fault signals, into semantic natural-language state descriptions and invokes an LLM to produce an interpretable intermediate reasoning trace. This reasoning identifies likely root causes, prunes the action space, and issues safe allocation decisions under policy constraints. Experiments on representative open-source microservice workloads show that ORACL improves root-cause identification accuracy by 15 percent, accelerates training by up to 24x, and improves quality of service by 6 percent in short-term scenarios, without deployment-specific retraining.

Cloud computing, despite its advantages in scalability, may not always fully satisfy the low-latency demands of emerging latency-sensitive pervasive applications. The cloud-edge continuum addresses this by integrating the responsiveness of edge resources with cloud scalability. Microservice Architecture (MSA) characterized by modular, loosely coupled services, aligns effectively with this continuum. However, the heterogeneous and dynamic computing resource poses significant challenges to the optimal placement of microservices. We propose REACH, a novel rescheduling algorithm that dynamically adapts microservice placement in real time using reinforcement learning to react to fluctuating resource availability, and performance variations across distributed infrastructures. Extensive experiments on a real-world testbed demonstrate that REACH reduces average end-to-end latency by 7.9%, 10%, and 8% across three benchmark MSA applications, while effectively mitigating latency fluctuations and spikes.

The transition from monolithic architecture to microservices has enhanced flexibility in application design and its scalable execution. This approach typically uses a computing cluster managed by a container orchestration platform to deploy microservices. However, this shift introduces significant challenges, particularly in the efficient scheduling of containerized services. These challenges are compounded by unpredictable scenarios such as dynamic incoming workloads with various execution traffic and variable communication delays among cluster nodes. Existing works often overlook the real-time traffic impacts of dynamic requests on running microservices, as well as the varied communication delays across cluster nodes. Consequently, even optimally deployed microservices could suffer from significant performance degradation over time. To address these issues, we propose a network and traffic-aware adaptive scheduling framework, TraDE, which can adaptively redeploy microservice instances to maintain desired performance amid changing traffic and network conditions within the hosting cluster. We have implemented TraDE as an extension to the Kubernetes platform. Additionally, we deployed realistic microservice applications in a real compute cluster and conducted extensive experiments to assess our framework's performance in various scenarios. The results demonstrate the effectiveness of TraDE in rescheduling running microservices to enhance end-to-end performance while maintaining a high goodput ratio. Compared with the existing method NetMARKS, TraDE outperforms it by reducing the average response time of the application by up to 48.3%, and improving the throughput by up to 1.2-1.5 x across workloads while maintaining a goodput ratio of 95.36%, and showing robust adaptive capability to meet QoS targets under sustained workloads and dynamic networking conditions.

The rapid growth of Internet of Medical Things (IoMT) devices has resulted in significant security risks, particularly the risk of malware attacks on resource-constrained devices. Conventional deep learning methods are impractical due to resource limitations, while Federated Learning (FL) suffers from high communication overhead and vulnerability to non-IID (heterogeneous) data. In this paper, we propose a split learning (SL) based framework for IoMT malware detection through image-based classification. By dividing the neural network training between the clients and an edge server, the framework reduces computational burden on resource-constrained clients while ensuring data privacy. We formulate a joint optimization problem that balances computation cost and communication efficiency by using a game-theoretic approach for attaining better training performance. Experimental evaluations show that the proposed framework outperforms popular FL methods in terms of accuracy, F1-score, high convergence speed, and less resource consumption. These results establish the potential of SL as a scalable and secure paradigm for next-generation IoMT security.

Circuit cutting lets a large quantum neural network (QNN) run as independent subcircuits on small devices, but rebuilding its outputs by reconstruction carries a classical sampling overhead exponential in the number of cuts - the dominant runtime cost in prior work. We ask whether, for machine-learning tasks, this step is necessary, and replace it with late fusion: each subcircuit is trained and measured independently, and a small classical head combines their outputs - a linear-cost, decision-level combination borrowed from multimodal learning. To characterize the trade-off we introduce a quantumness dial Q, a tunable reconstruction budget interpolating from pure fusion to full reconstruction, and a cut-entanglement diagnostic that indicates how much reconstruction a task needs (Spearman ρ=0.59 over 104 runs). Across synthetic and standard datasets, independently trained late fusion matches full reconstruction accuracy within 0.04 at every point of the controlled sweep and on every classical benchmark, at exponentially lower cost; it is also markedly more robust to shot and device noise. Controlled entangled-data experiments locate the boundary where fusion must fail. We do not claim advantage over classical machine learning - consistent with recent benchmarking, quantum offers no accuracy edge on these datasets. Late fusion is thus an efficient, noise-robust, self-characterizing alternative to reconstruction for circuit-cutting QML.

Distributed quantum computing (DQC) offers a practical path to scaling beyond the qubit and wiring limits of monolithic processors, but current compilation flows provide limited support for heterogeneous inter-chip communication. In multi-chip systems, remote operations are often slower, noisier, and less reliable than local gates, and their cost depends strongly on link quality, latency, and device calibration. As a result, compilation decisions such as circuit partitioning, link selection, and qubit mapping become tightly coupled, yet are often handled separately or with simplified communication models in existing solutions. We present DisMap, a calibration-aware distributed compiler that integrates the heterogeneous costs of both interand intra-chip operations into a unified hierarchical cost model. Given per-chip topology, noise calibration data, and inter-chip link specifications, DisMap constructs a global system topology, applies adaptive circuit partitioning to minimize expensive crosschip interactions, and performs iterative distributed qubit mapping to place logical operations on low-error qubits and route cross-chip communication over high-quality links. Evaluated on realistic multi-chip topologies inspired by IBM superconducting hardware, DisMap achieves up to 21.9% higher circuit fidelity and up to 92.6% lower execution cost compared to state-ofthe-art partitioning and mapping baselines, while maintaining practical compilation times across a wide range of circuit benchmarks.

The management of cloud resources for dynamic workloads presents a significant challenge. Learning-driven methods, including deep reinforcement learning, have emerged as promising solutions but necessitate extensive datasets for model training. Due to the practical constraints of acquiring real-world data, simulators are commonly employed to generate the requisite training workloads. However, a critical bottleneck arises from the architectural disconnect between CPU-bound simulation environments and GPU-accelerated training processes, leading to substantial data transfer overhead. To address this inefficiency, we propose a novel GPU-accelerated simulator implemented in the Julia programming language. Our framework is designed to unify simulation, model training, and inference, allowing these phases to execute almost entirely on GPUs. This approach effectively eliminates the traditional CPU-GPU communication bottleneck. Experimental evaluations demonstrate substantial performance gains over existing simulators, achieving up to a 96

Next-generation IoT applications increasingly span across autonomous administrative entities, necessitating silo-cooperative scheduling to leverage diverse computational resources while preserving data privacy. However, realizing efficient cooperation faces significant challenges arising from infrastructure heterogeneity, Non-IID workload shifts, and the inherent risks of adversarial environments. Existing approaches, relying predominantly on centralized coordination or independent learning, fail to address the incompatibility of state-action spaces across heterogeneous silos and lack robustness against malicious attacks. This paper proposes DeFRiS, a Decentralized Federated Reinforcement Learning framework for robust and scalable Silo-cooperative IoT application scheduling. DeFRiS integrates three synergistic innovations: (i) an action-space-agnostic policy utilizing candidate resource scoring to enable seamless knowledge transfer across heterogeneous silos; (ii) a silo-optimized local learning mechanism combining Generalized Advantage Estimation (GAE) with clipped policy updates to resolve sparse delayed reward challenges; and (iii) a Dual-Track Non-IID robust decentralized aggregation protocol leveraging gradient fingerprints for similarity-aware knowledge transfer and anomaly detection, and gradient tracking for optimization momentum. Extensive experiments on a distributed testbed with 20 heterogeneous silos and realistic IoT workloads demonstrate that DeFRiS significantly outperforms state-of-the-art baselines, reducing average response time by 6.4

The rapid proliferation of Internet of Things (IoT) devices across domains such as smart transportation, healthcare, and smart city infrastructure has intensified the demand for low-latency, energy-efficient, and scalable computing paradigms. While cloud computing has traditionally served as the backbone for IoT data processing, its inherent limitations have catalyzed the emergence of edge/fog computing, forming a distributed edge-cloud continuum. This review examines emerging trends in Artificial Intelligence (AI)-driven resource management within this continuum, with a focus on three directions: (1) the transition from centralized to distributed and collaborative intelligence, (2) cross-domain adaptation and knowledge transfer for heterogeneous IoT applications, and (3) the nascent integration of foundation models into edge environments. We ground our discussion in three complementary case studies: smart transportation as a representative vertical domain, cross-domain heterogeneous IoT application scheduling as a horizontal perspective, and industrial transferred arc plasma monitoring as an emerging industrial IoT scenario, and conclude with forward-looking research directions, including quantum-enhanced edge optimization, edge-native continual learning, digital twin-driven resource orchestration, and neuromorphic computing for ultra-low-power edge AI.

Recent stream computing systems have shifted from static, threshold-based straggler detection to adaptive, temporal-pattern-based prediction. Existing prediction approaches typically rely on autoregressive time-series analysis to model each node’s historical resource usage within a cluster. However, they often overlook cross-node state correlations, resulting in isolated prediction models that fail to capture the dynamic spatio-temporal dependencies among nodes. To address these limitations, we propose Ps-Stream, a proactive framework for straggler prediction and mitigation. First, Ps-Stream constructs a unified spatio-temporal graph by modeling compute nodes as a spatial graph at each time step and linking each node across adjacent time steps. Second, it learns spatio-temporal representations for straggler prediction by combining Graph Convolutional Networks (GCNs) to capture spatial dependencies with attention mechanisms to model temporal dynamics. Third, it incorporates a fine-grained task migration strategy that proactively reallocates tasks from predicted stragglers to the least-loaded nodes while considering task dependencies and migration costs to minimize migration overhead. Experimental results show that Ps-Stream improves straggler prediction accuracy by up to 26%, reduces maximum system latency by 62%, more than doubles maximum throughput, and improves resource utilization by up to 34% compared with state-of-the-art baselines.

High-performance computing (HPC) has evolved over decades through multiple architectural transitions, from vector supercomputers to massively parallel CPU clusters and GPU-accelerated systems, continuously expanding the frontier of scientific discovery. With the emergence of quantum processing units (QPUs) as practical computational accelerators, a new opportunity arises to further extend this trajectory by integrating quantum and classical computing paradigms. This paper presents Quantum Integrated High-Performance Computing (QHPC), a visionary architectural framework that unifies CPUs, GPUs, FPGAs, and QPUs as first-class heterogeneous resources. We propose a layered system design comprising unified resource management, quantum-aware scheduling, hybrid workflow orchestration, middleware and programming abstraction, interconnect technologies, and a tiered execution model enabling seamless workload partitioning across classical and quantum backends. A central aspect of our vision is a strong user requests abstraction layer that exposes heterogeneous resources through a unified job submission interface, similar in spirit to existing schedulers such as Slurm, allowing users to describe workloads in a consistent template independent of underlying compute type or location. Drawing insights from prior accelerator integration eras, we outline how QHPC can support emerging workloads in quantum chemistry, materials discovery, combinatorial optimization, and climate modeling. We conclude by highlighting open challenges in building scalable, reliable, and programmable quantum-classical infrastructures that seamlessly connect global users to heterogeneous compute resources for future quantum-classical HPC ecosystems.