Cold Start Latency in Serverless Computing: A Systematic Review, Taxonomy, and Future Directions

ACM Computing Surveys, Journal Year: 2024, Volume and Issue: 57(3), P. 1 - 36

Published: Oct. 17, 2024

Recently, academics and the corporate sector have paid attention to serverless computing, which enables dynamic scalability an economic model. In users only pay for time they actually use resources, enabling zero scaling optimise cost resource utilisation. However, this approach also introduces cold start problem. Researchers developed various solutions address problem, yet it remains unresolved research area. article, we propose a systematic literature review on latency in computing. Furthermore, create detailed taxonomy of approaches latency, investigate existing techniques reducing frequency. We classified current studies into several categories such as caching application-level optimisation-based solutions, well Artificial Intelligence/Machine Learning-based solutions. Moreover, analyzed impact quality service, explored mitigation methods, datasets, implementation platforms, them based their common characteristics features. Finally, outline open challenges highlight possible future directions.

Language: Английский

---

The role of cognitive computing in NLP

Frontiers in Computer Science, Journal Year: 2025, Volume and Issue: 6

Published: Jan. 10, 2025

The integration of Cognitive Computing and Natural Language Processing (NLP) represents a revolutionary development Artificial Intelligence, allowing the creation systems capable learning, reasoning, communicating with people in natural meaningful way. This article explores convergence these technologies highlights how they combine to form intelligent understanding interpreting human language. A comprehensive taxonomy NLP is presented, which classifies key tools techniques that improve machine language generation. also practical applications, particular, accessibility for visual impairments using advanced Intelligence-based tools, as well analyze political discourse on social networks, where provide insight into public sentiment information dynamics. Despite significant achievements, several challenges persist. Ethical concerns, including biases AI, data privacy societal impact, are critical address responsible deployment. complexity poses interpretative challenges, while multimodal real-world deployment difficulties impact model performance scalability. Future directions proposed overcome through improved robustness, generalization, explainability models, enhanced scalable, resource-efficient thus provides view current advancements outlines roadmap inclusive future NLP.

Language: Английский

---

Beyond the Cloud: Federated Learning and Edge AI for the Next Decade

Journal of Computer and Communications, Journal Year: 2025, Volume and Issue: 13(02), P. 37 - 50

Published: Jan. 1, 2025

Language: Английский

---

Edge and Cloud Computing in Smart Cities

Future Internet, Journal Year: 2025, Volume and Issue: 17(3), P. 118 - 118

Published: March 6, 2025

The evolution of smart cities is intrinsically linked to advancements in computing paradigms that support real-time data processing, intelligent decision-making, and efficient resource utilization. Edge cloud have emerged as fundamental pillars enable scalable, distributed, latency-aware services urban environments. Cloud provides extensive computational capabilities centralized storage, whereas edge ensures localized processing mitigate network congestion latency. This survey presents an in-depth analysis the integration cities, highlighting architectural frameworks, enabling technologies, application domains, key research challenges. study examines allocation strategies, analytics, security considerations, emphasizing synergies trade-offs between paradigms. present also notes future directions address critical challenges, paving way for sustainable development.

Language: Английский

---

Health is beyond genetics: on the integration of lifestyle and environment in real-time for hyper-personalized medicine

Frontiers in Public Health, Journal Year: 2025, Volume and Issue: 12

Published: Jan. 7, 2025

Hyper-personalized medicine represents the cutting edge of healthcare, which aims to tailor treatment and prevention strategies uniquely each individual. Unlike traditional approaches, often adopt a one-size-fits-all or even broadly personalized approach based on broad genetic categories, hyper-personalized considers an individual's comprehensive health data by integrating unique biological, genetic, lifestyle, environmental influences. This method goes beyond simple profiling recognizing that outcomes are influenced complex interactions among our environment, daily routines, physiological processes responses.Central is integration lifestyle factors. Lifestyle habits, such as diet (Dalwood et al., 2020; Genel Marx Hepsomali & Groeger, 2021; Dinu 2022; Yang Sadler 2024), exercise (Chow Qiu Ross D'Onofrio 2023; Isath Mahindru Ashcroft 2024; Ponzano sleep patterns (Hepsomali Baranwal Eshera Lim Sletten Uccella, Weinberger 2023), directly impact health. Hence, understanding these factors helps interventions align with day-to-day realities Environmental factors, air quality (Cheek Markandeya Shukla Tang Abdul-Rahman Bedi Bhattacharya, climate (Coates Ebi Helldén Reismann Rocque Zhang Münzel Palmeiro-Silva exposure pollutants (Qadri Faiq, 2019; Petroni Lin Sun Xu Yu Levin Shetty Deziel Villanueva Sharma also play significant roles in determining outcomes. By continuously monitoring analyzing elements, healthcare providers can create dynamic plans adapt real-time changes. would allow for proactive measures optimized care.To enable model care, advanced technologies like quantum computing, artificial general intelligence (AGI), internet things (IoT), 6G connectivity crucial roles. Quantum computing offers ability process vast intricate datasets, those required between markers, exposures, choices, far greater speed accuracy than classical (Munshi Kumar Stefano, Ullah Garcia-Zapirain, 2024). AGI, its adaptive learning capabilities, analyze make sense this provide precise, evolving recommendations change patient's environment does (Liu Mitchell, Tu IoT devices, including wearables sensors, gather continuous from individuals, tracking physical activity, biometrics, conditions humidity (Puri Islam Mathkor Rocha Šajnović Salam, With advent connectivity, seamlessly transferred processed real time, enabling instant feedback intervention (Nayak Patgiri, Nguyen Ahad Kumar, Kaur, Mahmood Mihovska 2024).Together, form backbone model, will push medical practices highly responsive, individual-centered As advancements continue evolve, has potential fundamentally reshape offering truly support long-term well-being.

Language: Английский

---

Characterizing Perception Deep Learning Algorithms and Applications for Vehicular Edge Computing

Algorithms, Journal Year: 2025, Volume and Issue: 18(1), P. 31 - 31

Published: Jan. 8, 2025

Vehicular edge computing relies on the computational capabilities of interconnected devices to manage incoming requests from vehicles. This offloading process enhances speed and efficiency data handling, ultimately boosting safety, performance, reliability connected While previous studies have concentrated processor characteristics, they often overlook significance connecting components. Limited memory storage resources pose challenges, particularly in context deep learning, where these limitations can significantly affect performance. The impact contention has not been thoroughly explored, especially regarding perception-based tasks. In our analysis, we identified three distinct behaviors contention, each interacting differently with other resources. Additionally, investigation Deep Neural Network (DNN) layers revealed that certain convolutional experienced computation time increases exceeding 2849%, while activation showed a rise 1173.34%. Through characterization efforts, model workload behavior according their configuration demands allows us quantify effects contention. To knowledge, this study is first characterize influence vehicular workloads, strong emphasis dynamics DNN layers.

Language: Английский

---

Autonomic Cloud Computing: Research Perspective

Published: Jan. 1, 2025

Language: Английский

---

Deep Reinforcement Learning-Enabled Computation Offloading: A Novel Framework to Energy Optimization and Security-Aware in Vehicular Edge-Cloud Computing Networks

Sensors, Journal Year: 2025, Volume and Issue: 25(7), P. 2039 - 2039

Published: March 25, 2025

The Vehicular Edge-Cloud Computing (VECC) paradigm has gained traction as a promising solution to mitigate the computational constraints through offloading resource-intensive tasks distributed edge and cloud networks. However, conventional computation mechanisms frequently induce network congestion service delays, stemming from uneven workload distribution across spatial Roadside Units (RSUs). Moreover, ensuring data security optimizing energy usage within this framework remain significant challenges. To end, study introduces deep reinforcement learning-enabled for multi-tier VECC First, dynamic load-balancing algorithm is developed optimize balance among RSUs, incorporating real-time analysis of heterogeneous parameters, including RSU load, channel capacity, proximity-based latency. Additionally, alleviate in static deployments, proposes deploying UAVs high-density zones, dynamically augmenting both storage processing resources. an Advanced Encryption Standard (AES)-based mechanism, secured with one-time encryption key generation, implemented fortify confidentiality during transmissions. Further, context-aware caching strategy preemptively store processed tasks, reducing redundant computations associated overheads. Subsequently, mixed-integer optimization model formulated that simultaneously minimizes consumption guarantees latency constraint. Given combinatorial complexity large-scale vehicular networks, equivalent learning form given. Then learning-based designed learn close-optimal solutions under conditions. Empirical evaluations demonstrate proposed significantly outperforms existing benchmark techniques terms savings. These results underscore framework's efficacy advancing sustainable, secure, scalable intelligent transportation systems.

Language: Английский

---

Edge AI Deploying Artificial Intelligence Models on Edge Devices for Real-Time Analytics

ITM Web of Conferences, Journal Year: 2025, Volume and Issue: 76, P. 01009 - 01009

Published: Jan. 1, 2025

Because of its on-the-go nature, edge AI has gained popularity, allowing for realtime analytics by deploying artificial intelligence models onto devices. Despite the promise Edge evidenced existing research, there are still significant barriers to widespread adoption with issues such as scalability, energy efficiency, security, and reduced model explainability representing common challenges. Hence, while this paper solves in a number ways, real use case deployment, modular adaptability, dynamic specialization. Our paradigm achieves low latency, better security efficiency using light-weight models, federated learning, Explainable (XAI) smart edge-cloud orchestration. This framework could enable generic beyond specific applications that depend on multi-modal data processing, which contributes generalization across various industries healthcare, autonomous systems, cities, cybersecurity. Moreover, work will help deploy sustainable employing green computing techniques detect anomalies near real-time critical domains helping ease challenges modern world.

Language: Английский

---

HealthEdgeAI: GAI and XAI Based Healthcare System for Sustainable Edge AI and Cloud Computing Environments

Concurrency and Computation Practice and Experience, Journal Year: 2025, Volume and Issue: 37(9-11)

Published: April 10, 2025

ABSTRACT Coronary heart disease is a leading cause of mortality worldwide. Although no cure exists for this condition, appropriate treatment and timely intervention can effectively manage its symptoms reduce the risk complications such as attacks. Prior studies have mostly relied on limited dataset from UC Irvine Machine Learning Repository, predominantly focusing (ML) models without incorporating Explainable Artificial Intelligence (XAI) or Generative (GAI) techniques enhancement. While some research has explored cloud‐based deployments, implementation edge AI in domain remains largely under‐explored. Therefore, paper proposes HealthEdgeAI , sustainable approach to prediction that enhances XAI through GAI‐driven data augmentation. In our research, we assessed multiple by evaluating accuracy, precision, recall, F1‐score, area under curve (AUC). We also developed web application using Streamlit demonstrate methods employed FastAPI serve optimal model an API. Additionally, examined performance these cloud computing settings comparing key Quality Service (QoS) parameters, average response rate throughput. To highlight potential computing, tested devices with both low‐ high‐end configurations illustrate differences QoS. Ultimately, study identifies current limitations outlines prospective directions future AI‐based environments.

Language: Английский

---