Hybrid profitability-aware PSO model with dissimilarity index for cost-efficient and balanced cloud resource allocation

Identifying confounding variables is fundamental for robust observational studies, yet the traditional manual process is a time-consuming and subjective barrier for researchers. Recent advances in Retrieval-Augmented Generation (RAG) offer a promising solution, but most existing systems rely on full-text access, cloud-hosted APIs, or manually curated knowledge graphs, raising concerns about privacy, copyright, and computational cost, and making local deployment difficult. This study developed and evaluated a heuristic tool to scope candidate confounders for adjustment in observational studies. Using a locally deployed, abstract-only RAG architecture, our tool generates a traceable shortlist of candidate confounders from PICO (Population, Intervention, Comparison, Outcome) queries over medical abstracts. We implemented a three-stage architecture for PICO-based scoping of candidate confounder. The pipeline was deployed on an all-in-one local server and evaluated using 1,000 expert-curated PICO queries spanning 20 clinical specialties. Performance was assessed along four dimensions-internal consistency, output volume, efficiency, and clinical acceptance-by a multi-institutional clinician panel, and was compared with a graph-only SemMedDB baseline. Across repeated runs, the pipeline showed high internal consistency (candidate confounder list consistency 94.6%±8.7%; PMID set consistency 79.4%±23.5%). It suggested a median of 6 candidate confounders (IQR 8) for adjustment and retrieved a median of 33 unique PMIDs (IQR 7) per query. Median processing time was 44.50s (IQR 31.72). Expert review yielded an overall clinical acceptance rate of 87.12%. In an exploratory capacity, a locally deployed, abstract-only RAG workflow can generate clinically interpretable and traceable candidate confounder suggestions to support early-stage observational study design, particularly in settings with privacy constraints or limited access to full texts and cloud resources. NA.

In the era of modern technologies, introduced the widespread use of cloud computing and other solutions that revolutionized the storage and management of information. Cloud-based network threat identification and risk management using applying Log Analysis and Machine learning is intended to recognize, evaluate, and analyze cybersecurity risks in contemporary cloud settings. As cloud computing becomes more widely used, dynamic, expansive, and dispersed network infrastructures cannot be managed by conventional perimeter-based security measures. The intelligent threat detection system proposed in this research gathers and examines system and network logs produced by cloud resources in order to instantly spot malicious activity. In order to classify network behavior as either normal or abnormal, the model uses machine learning algorithms to extract pertinent data such traffic patterns, protocol usage, access frequency, and temporal behavior. In order to help security teams, prioritize incidents, risk assessment is carried out by allocating weighted scores based on threat severity, asset criticality, and previous behavior. Automated analysis, scalable log ingestion, and visual dashboards for tracking risks and threats are all supported by the architecture. By combining machine learning based random forest detection with log analysis, the suggested approach increases reaction efficiency, decreases false positives, and increases threat visibility. This project shows a realistic, affordable, and expandable solution to cloud security, which makes it appropriate for both real-world deployment scenarios and academic demonstrations. This technology has the potential to revolutionize our approach to cybersecurity and system security.

This paper presents a digital-twin-based model predictive control framework for autonomous process control, demonstrated in a virtual experiment on thermal food processing in a convection oven. In combination with prior work, this approach enables simulation-centered food scientists to deploy physics-based simulation models in live process control environments. The digital twin is realized as a physics-based, data-driven reduced-order model (ROM) that provides faster-than-real-time predictions. The ROM is trained on trajectories from a high-fidelity multiphysics finite-element model of chicken fillets. A central contribution is a model predictive control scheme that overcomes the common fixed-initial-condition limitation of augmented neural ordinary differential equation ROMs: a dedicated sub-optimization step re-synchronizes the surrogate to the measured state of the food item at each control instant, allowing reliable live re-optimization without access to internal ROM states. The controller optimizes oven temperature setpoints to meet target food-quality metrics (core temperature, moisture content, texture) and autonomously accommodates changes to the planned end time during operation. Quantitatively, the ROM achieves relative time-series errors of 0.18–0.49%, and the control algorithm evaluates 501 trajectories of 1800 s real time in a total of 46.6 s on a single core of a processor, demonstrating on-device feasibility without cloud or edge resources. Receding-horizon model predictive control of the remaining setpoints mitigates model–reality mismatch, enforces user-defined food metrics, and sustains closed-loop performance under autonomous operation. • Software-agnostic pipeline to deploy digital-twin surrogates for online control. • Accurate, faster-than-real-time predictions enable autonomous operation. • Online sub-optimization re-synchronizes predictions to measured core temperature. • Level 5 autonomy case beyond the trained time window. • On-device digital twin operation possible without cloud or edge resources.

The modernization of logistics and distribution has promoted the efficiency improvement of tobacco inventory scheduling. This study provides data reference for intelligent transportation scheduling by establishing a real-time tobacco delivery model, comprehensively considering the tobacco inventory in each period, and calculating a more accurate real-time delivery demand. A single source point non-full load type time-limited distribution model is constructed, and the optimal path selection of transportation vehicles is abstracted as a vehicle total distance minimization solving problem. The C-W savings solving algorithm is introduced, combined with the soft time window cost calculation of each vehicle waiting, to correct the cost saving value and determine the optimal scheduling route. After the optimization of C-W saving solution algorithm, the shortest path between demand points can be shortened to about 35.64km, and the saving value is improved to more than 100, which can reasonably make full use of the transportation resources to complete the real-time distribution of tobacco.

The rapid proliferation of computing technologies, cloud infrastructure, and Internet of Things (IoT) devices has intensified global energy consumption, accelerated electronic waste generation, and amplified greenhouse gas emissions. These trends pose acute sustainability challenges, particularly for developing economies in Sub-Saharan Africa where power infrastructure remains unreliable, e-waste governance frameworks are nascent, and financial resources for green technology adoption are severely constrained. This paper presents a comprehensive review of green computing techniques and examines their potential to create environmentally sustainable computing environments in developing economies. Drawing on a structured review of peer-reviewed literature published between 2010 and 2024, the study analyses key green computing dimensions including energy-efficient data center design, server virtualization, cloud resource optimization, e-waste lifecycle management, green procurement, AI-driven power management, and renewable energy integration. The findings reveal that, while significant adoption gaps persist in low-income settings, targeted policy interventions, capacity-building programmes, and south-south technology transfer partnerships can substantially advance green computing uptake. The study concludes that green computing is not a luxury exclusive to technologically advanced nations but a strategic imperative for developing economies seeking to leapfrog legacy, energy-intensive ICT infrastructure and achieve sustainable digital development.

ABSTRACT Cloud infrastructure governance increasingly depends on the ability to enforce measurable service-level objectives (SLOs)—the contractual commitments that underpin availability, performance, and operational resilience in modern enterprise environments. When SLOs are violated, organizations face compounding consequences: service disruptions, compliance exposure, audit findings, and degraded trust. Predictive autoscaling offers a proactive control mechanism, but its governance value is realized only when capacity decisions are grounded in calibrated, risk-aware uncertainty rather than deterministic point forecasts. This article introduces SLO-Aware Heteroscedastic LSTM (SLOAH-LSTM), a probabilistic workload forecasting model that embeds asymmetric risk weighting into its learning objective, directing forecast precision where it matters most for compliance. To support auditable and explainable capacity decisions, we evaluate three interpretable decision mappings (mean, quantile, and cost-aware Monte Carlo selection). A simulation framework replicating realistic cloud controller constraints demonstrates our approach reduces SLO violation rates from ≈6.5 percent to ≈3.0 percent with only a moderate increase in provisioned resources (avg pods ≈20 → 25) and a markedly improved combined cost index (≈55 vs. ≈85). These findings establish a practical, low-overhead framework for governing cloud resource allocation with quantifiable compliance guarantees, directly supporting IT audit objectives around availability assurance, risk-aware capacity planning, and the operational integrity of cloud-native enterprise infrastructure.

Self-supervised multi-scale cloud workload prediction with time series data augmentation.

Adaptive and energy-efficient computing infrastructures that can manage diverse workloads are essential for the integration of renew- able energy into smart grids. Predictive accuracy and global optimization in dynamic energy environments are limited by the use of heuris- tic scheduling techniques or lightweight artificial intelligence in existing load balancing frameworks. In order to improve load balancing in hy- brid edge-cloud renewable energy networks, this paper suggests a novel framework that combines Convolutional Neural Networks (CNNs) with Brown Bear Optimization (BBO). In order to produce precise short-term predictions of energy availability and task demands, the CNN module ex- tracts deep spatiotemporal features from solar irradiance, weather data, and workload traces. A BBO-driven scheduler uses these forecasts to minimize a multi-objective cost function that balances latency, energy consumption, and task completion reliability. Priority-aware scheduling and intelligent task migration between edge and cloud resources are en- sured by the dynamic classification of tasks into three categories: Criti- cal Real-Time, Latency-Sensitive, and Delay-Non-Critical. In comparison to LSTM-based and heuristic approaches, simulation studies using real- world solar and grid workload datasets show that the suggested CNN- BBO framework improves task completion rates to 98.6%, maintains low latency, and reduces average energy consumption by over 35%. The out- comes demonstrate that CNN-BBO integration offers next-generation smart energy infrastructures a scalable, resilient, and sustainable solu- tion.

Conditional skip liquid neural networks: an efficient inference framework for large-scale time-series cloud resource prediction

Abstract—Cloud computing allows access to computing re-sources in a scalable and on-demand fashion, yet effective, resource allocation is also a major problem because of the dynamic workloads and unpredictable user demands. The com-mon allocation approaches, including the provisioning that is not dynamic and scheduling that is based on the rules, tend to lead to low utilization of resources, high latency, and high operational costs. To solve these problems, this paper suggests an AI-based solution to optimize the resources distribution in the cloud environment. The suggested system will use the methods of Artificial Intelligence, such as, but not limited to, Machine Learning, Deep Learning, and Reinforcement Learning, to predict the workload trends and dynamically distribute the resources. The predictive models are trained using historical data like CPU usage, memory consumption, and network bandwidth. These models help to pre-dict resource needs with accuracy and proactively and efficiently assign them. Reinforcement Learning also boosts decision-making as it continuously improves allocation policies through system feedback. Experimental findings show that the suggested solution can be used to better utilize resources and minimize latency and the overall system performance than conventional solutions. Scalability and adaptability in a large-scale cloud environment is also supported by the model. This paper has shown that the application of AI methods in cloud resource management is an efficient way to get effective, cost-efficient, and smart resource allocation. Keywords: Cloud Computing, Dynamic Resource Allocation, AI-based Scheduling, Reinforcement Learning, LSTM, Predictive Analytics, Virtualization, Quality of Service (QoS). Index Terms—component, formatting, style, styling, insert

Machine learning (ML)-enabled systems like autonomous driving systems (ADSs) face challenges meeting safety and performance requirements in diverse environments, especially in resource-constrained, latency-sensitive edge-cloud settings. These challenges often arise from the ML models’ limitations, including poor generalization to unseen conditions. Static ML models often struggle to generalize to unseen scenarios, particularly under the latency and resource constraints of edge-cloud infrastructure. Adaptive algorithms using ML system switching have been proposed, but existing approaches frequently lack generalizability, support for common black-box systems, and effective use of distributed edge-cloud resources. This paper presents a novel adaptive ML-enabled edge-cloud system framework to address these shortcomings. Our framework combines cloud-based pre-runtime analysis, which leverages simulation for behavioral understanding and scenario-to-system mapping, with collaborative edge-cloud runtime adaptation featuring dynamic ML model switching. It supports black-box systems and aims to balance safety and efficiency by utilizing appropriate edge and cloud resources situationally. Preliminary CARLA-based evaluation of the edge runtime component suggests our framework can potentially improve the safety-efficiency trade-off compared to single-model ADSs in some scenarios. Moreover, extensive experiments using the MetaDrive simulator with 100,000 randomized driving scenarios demonstrate that the adaptive system improves safety by 2.6% while doubling computational efficiency compared to a single-model baseline. These results validate the framework’s scalability and the feasibility of data-driven scenario–system mapping for adaptive ML-enabled autonomous systems operating across edge and cloud environments.

Effective resource allocation in cloud computing continues a critical challenge due to dynamic loads, stringent service-level expectations, and the need to balance execution time, energy, and cost. This study suggests a hybrid framework that integrates Deep Q-Learning (DQL) with Particle Swarm Optimization (PSO) to aid adaptive, multi-objective scheduling. DQL learns allocation strategies through interaction with the cloud environment, while PSO performs global search to refine action selection and accelerate convergence. Using Cloud Sim with real and synthetic workloads (Google Cluster, Planet Lab traces), the proposed method achieved a 35% reduction in average task execution time (from 245s to 159s) and a ~ 40% relative growth in resource utilization (from 60.1% to 84.6%), reduced SLA violations from 28 to 8, and lowered energy consumption to 6.3 kWh, outperforming standalone and hybrid models across 30 independent runs. Statistical tests (two-tailed t-test, α = 0.05) confirm significance. These results demonstrate that coupling reinforcement learning among swarm intelligence yields adaptive, high-quality decisions on behalf of real-time cloud resource scheduling.

The Compute Continuum—spanning IoT, Edge, Cloud, and HPC resources—is reshaping how hyper-distributed applications are designed and orchestrated. Traditional service orchestrators and workload management systems rely on centralized runtimes; however, the emerging paradigm requires decentralized coordination, where autonomous agents cooperate to achieve common goals and dynamically distribute workloads. Consensus algorithms play a crucial role in multi-agent systems (MAS), as they enable agents to reach agreement on how to coordinate and execute functionalities in a cooperative manner. While consensus has previously been applied to distributed job selection, here we extend its use to swarm environments. In this setting, agents autonomously decide which service functionalities (i.e., roles) to execute based on their capabilities and the real-time quality of service (QoS). Functionalities can be elastically activated or terminated as application needs evolve. To support this model, we leverage the COLMENA framework, a programming environment for defining and managing such dynamic services. We apply a greedy consensus-based approach to modern power systems, which are increasingly decentralized due to the large-scale integration of renewable energy sources. Centralized power plants are giving way to distributed, intermittent resources that require decentralized control paradigms. To demonstrate this, we simulate the Northeastern Power Coordinating Council’s (NPCC) 140-bus grid using the ANDES simulator in conjunction with the COLMENA middleware. We deploy this use case across six different sites in the FABRIC testbed, using up to 60 different nodes. Our results show that, under contingency scenarios such as load and generator disconnections, agents self-organize, elect local leaders, and execute optimization algorithms to stabilize grid frequency. Detection and organization times remain below 10s across all experiments, even as the number of agents per area scales from 3 to 10. Stability is restored within approximately 27s and 40s for the respective cases. Resource overhead is minimal, with CPU and memory usage remaining below 7.5% and 2%, respectively. Experiment automation and reproducibility are ensured through Kiso. These findings indicate that role-based programming models complement traditional workflows and that consensus-driven coordination can effectively decentralize decision-making in swarm environments. This approach represents a step toward enabling resilient, decentralized power systems.

The integration of artificial intelligence (AI) and machine learning (ML) into cloud computing has led to the emergence of distributed AI/ML applications operating across the cloud–edge continuum. However, privacy concerns, regulatory compliance, and ethical constraints present challenges in determining optimal workload distribution. This paper explores a strategic allocation framework that dynamically assigns AI/ML workloads between cloud and edge resources based on data sensitivity, computational efficiency, and policy requirements. By leveraging intelligent orchestration mechanisms, we propose an adaptive approach that enhances performance, ensures compliance with regulatory frameworks, and upholds ethical AI principles. Our findings contribute to the development of secure, efficient, and responsible AI/ML deployment in cloud-edge environments.

The Internet of Vehicles (IoV) enables data exchange among individuals, cloud resources, road infrastructures, and vehicles, interconnected through Vehicular Ad Hoc Networks (VANETs). VANETs comprise vehicles with Onboard Units (OBUs), Roadside Units (RSUs), and a Trusted Party Agent (TPA). The data transmission among these entities supports seamless interaction and collaborative traffic management. However, data transmission on public communication channels in VANETs presents significant challenges, including security, privacy, and authentication of participating entities. Although numerous key exchange and authentication protocols have been introduced to tackle these issues, many protocols remain vulnerable to various attacks, such as a vehicle, RSU, TPA impersonation, denial of service, physical cloning, and desynchronization attacks. Therefore, to address these vulnerabilities, we propose a key exchange protocol that leverages hash functions and Advanced Encryption Standard (AES) encryption. Our protocol also integrates the Physical Unclonable Function (PUF), enhancing its resistance to physical or cloning attacks. Additionally, it effectively counters threats like impersonation, session key leakage, ephemeral secret leakage, and desynchronization attacks. We validate the security and reliability of our protocol through both formal and informal analysis. Informal analysis highlights the protocol’s essential security features, while formal analysis provides robust substantiation. Performance evaluation reveals that our protocol achieves an average reduction of 35.53%, and 53.77%, in communication and computation overheads.

This research presents a multi-objective, energy-aware workflow scheduling framework for heterogeneous cloud–edge environments that addresses both efficiency and data integrity challenges. Conventional encryption-based security mechanisms, although effective in protecting data during task offloading, often introduce significant computational and communication overhead, leading to degraded system performance. To overcome this limitation, this work proposes the consensus security-integrity and quality-aware workflow scheduler (CSIQA-WS), which integrates energy-aware scheduling with a lightweight, consensus-driven security mechanism. The model incorporates automatic service management and an attack prevention module to detect and mitigate malicious behavior during inter-node data transmission while maintaining quality of service (QoS) constraints. A dynamic coordination between edge and cloud resources enables efficient workload distribution and robust resource utilization. Experimental evaluation using scientific workflow benchmarks demonstrates that CSIQA-WS significantly reduces processing time and energy consumption compared to existing approaches. The proposed model achieves up to 92.29% reduction in processing time and consistently improves overall QoS while preserving data integrity in dynamic execution environments. These results indicate that CSIQA-WS provides an effective and scalable solution for secure and energy-efficient workflow scheduling in modern cloud–edge systems.

Agentic AI has emerged as a promising paradigm for autonomous reasoning and execution in complex AI-driven applications; however, its effective deployment in cloud-native environments remains challenging due to the lack of unified platform architectures that jointly support task decomposition, multi-agent collaboration, and adaptive cloud resource orchestration. In practical scenarios such as automated data analytics, AI DevOps, and MLOps pipelines, Agentic AI systems must operate over dynamic containerized infrastructures where resource availability, execution cost, and failure conditions continuously change. Existing approaches typically decouple agent-level decision making from cloud-native scheduling, resulting in limited scalability and poor robustness. To address these limitations, this paper proposes CANAO, a Cloud-Aware Native Agentic AI framework for adaptive task orchestration in cloud-native environments. CANAO models complex AI workloads as dynamically reconfigurable task dependency graphs and enables coordinated collaboration among Planner, Executor, and Critic agents. By incorporating real-time cloud resource awareness into the agent orchestration loop, CANAO supports adaptive scheduling, partial task re-planning, and self-healing execution on Kubernetes-based platforms. A prototype system is implemented using cloud-native technologies and evaluated on representative automated data analysis and AI DevOps workflows. Experimental results show that CANAO significantly outperforms baseline orchestration methods under dynamic cloud conditions. Compared with static DAG-based scheduling, CANAO reduces end-to-end task execution time by approximately 34.3% and cloud resource cost by nearly 30%, while lowering the task failure rate by over 34%. These improvements demonstrate the effectiveness of cloud-aware agent collaboration and adaptive task orchestration in large-scale cloud-native AI workflows.

The rapid and unprecedented growth of urban populations has led to increased traffic congestion and road accidents, necessitating the development of Intelligent Transportation Systems (ITS) to improve road safety, traffic management, and driving efficiency. Among emerging technologies, Vehicular Ad Hoc Networks (VANETs) integrated with cloud computing have gained significant attention for enabling seamless vehicle-to-vehicle (V2V) and vehicle-to-cloud (V2C) communication. In such systems, vehicles are equipped with advanced sensors that collect real-time data about the environment, roadway, and surrounding vehicles. This data is processed, stored, and shared through cloud infrastructure to support a range of applications, including real-time traffic updates, collision avoidance alerts, navigation assistance, infotainment(“information” and “entertainment.”), and emergency response. The VAN-Cloud architecture offers a scalable and efficient framework for managing vehicular data and cloud resources, ensuring secure communication, low latency, and reliable data exchange. Furthermore, modern vehicles are increasingly equipped with intelligent onboard units that enhance these capabilities, contributing to the development of safer, smarter, and more efficient transportation systems.

Cloud computing has a major impact on the IT industry. How to price and allocate cloud resources to meet users’ requirements is an important problem. This paper proposes a dynamic mechanism to pricing cloud services, which can work in complex environments such as distributed system and uncertain budget constraints. A direct relationship between QoS and price is established. The approach uses an optimization technique to estimate the potential transaction price in the distributed network. It can allocate cloud resources under uncertainties, where providers can optimize their revenues, and consumers can obtain the resources at a relatively low price.