Article
Computer Science, Information Systems
Zaakki Ahamed, Maher Khemakhem, Fathy Eassa, Fawaz Alsolami, Abdullah S. Al-Malaise Al-Ghamdi
Summary: Proactive resource management in Cloud Services is important for cost effectiveness and addressing issues such as SLA violations and resource provisioning. Workload prediction using Deep Learning (DL) is popular for analyzing cloud environment data, but the quality of the training data influences the model's performance. Existing works in this domain often lack uniformity in data sources, leading to decreased efficacy of DL models. In this study, DL models are used to analyze real-world workloads from SWF, and the LSTM model exhibits the best performance. The paper also addresses the lack of literature on DL in workload prediction in cloud computing environments.
Article
Computer Science, Information Systems
Piotr Nawrocki, Patryk Osypanka, Beata Posluszny
Summary: Predicting and managing computing resource usage in cloud computing is important for cost optimization. This paper presents a novel approach that combines data-driven adaptation of prediction algorithms to generate accurate short- and long-term cloud resource usage predictions. The proposed solution outperforms static algorithm selection and achieves better prediction quality, reducing costs by up to 80.68%.
JOURNAL OF GRID COMPUTING
(2023)
Article
Computer Science, Information Systems
Ali Shahidinejad, Mostafa Ghobaei-Arani, Mohammad Masdari
Summary: This paper presents a hybrid solution to handle resource provisioning issue in cloud environment by using ICA and K-means for workload clustering, and decision tree algorithm for scaling decisions. The study shows that the proposed approach significantly reduces total cost and response time, while increasing CPU utilization and elasticity.
CLUSTER COMPUTING-THE JOURNAL OF NETWORKS SOFTWARE TOOLS AND APPLICATIONS
(2021)
Article
Computer Science, Information Systems
Shihong Hu, Weisong Shi, Guanghui Li
Summary: This article introduces two major challenges faced in the real deployment of containers at edge servers: the varying workload of service requests and the startup delay of containers. To address these challenges, a containerized edge computing framework called CEC is proposed, which focuses on the smart connected community with multiple intelligent applications. CEC integrates workload prediction and resource pre-provisioning to achieve low latency and high utilization of edge resources for user service requests.
IEEE TRANSACTIONS ON MOBILE COMPUTING
(2023)
Article
Computer Science, Information Systems
Bowen Fei, Xiaomin Zhu, Daqian Liu, Junjie Chen, Weidong Bao, Ling Liu
Summary: This article proposes a method of elastic resource provisioning using data clustering in a cloud service platform. The method effectively meets the demands of different types of tasks through tasks clustering, prediction of the amount of tasks, and dynamic resource provisioning.
IEEE TRANSACTIONS ON SERVICES COMPUTING
(2022)
Article
Computer Science, Information Systems
Wiem Matoussi, Tarek Hamrouni
Summary: The proposed method aims to predict the number of requests at a SaaS service to achieve precise forecasting results and optimized response time, striking a balance between execution time and prediction accuracy.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2022)
Article
Computer Science, Information Systems
Patryk Osypanka, Piotr Nawrocki
Summary: Cloud computing services are increasingly popular, but they can also lead to high operating costs. Many efforts have been made to optimize cloud resource usage and reduce expenses, but such optimization often comes at the expense of service responsiveness and quality, especially when dealing with real-world data and anomalies. This article proposes a novel approach that incorporates machine learning-based load prediction, discovery of service characteristics, long-term resource planning, anomaly detection, and continuous monitoring to achieve cost optimization without sacrificing performance. Evaluation using Microsoft's Azure cloud environment showed cost savings ranging from 31% to 89% depending on the test scenario.
IEEE TRANSACTIONS ON SERVICES COMPUTING
(2023)
Article
Computer Science, Hardware & Architecture
Sounak Banerjee, Sarbani Roy, Sunirmal Khatua
Summary: The demand for cloud-based services is rapidly increasing due to their scalability and cost-effectiveness. As a result, the size and maintenance cost of data centers are growing, making it crucial to develop a proper resource management plan. The proposed machine learning-based workload prediction approach in this paper improves resource utilization and reduces overall energy consumption in data centers.
JOURNAL OF SUPERCOMPUTING
(2021)
Article
Computer Science, Information Systems
Shivani Singh, Razia Sulthana, Tanvi Shewale, Vinay Chamola, Abderrahim Benslimane, Biplab Sikdar
Summary: Edge computing is a technological advancement that connects sensors and provides services at the device end, but security is a major concern. This article explores the security and privacy issues in different layers of the Edge computing architecture and the machine learning algorithms used to address these concerns. It also discusses various types of attacks on the Edge network and introduces intrusion detection systems and machine learning algorithms that overcome these security and privacy challenges.
IEEE INTERNET OF THINGS JOURNAL
(2022)
Article
Computer Science, Hardware & Architecture
Javad Dogani, Farshad Khunjush, Mohammad Reza Mahmoudi, Mehdi Seydali
Summary: This paper presents a hybrid method for predicting multivariate time series workload of host machines in cloud data centers. It constructs a training set through statistical analysis, extracts hidden spatial features using convolutional neural networks, and extracts temporal correlation features using a GRU network optimized with attention mechanism.
JOURNAL OF SUPERCOMPUTING
(2023)
Article
Computer Science, Theory & Methods
Binlei Cai, Bin Wang, Meihong Yang, Qin Guo
Summary: AutoMan is a learning-driven resource manager for microservices that uses a multi-agent deep deterministic policy gradient (MADDPG) method to efficiently allocate resources while guaranteeing the end-to-end tail latency Service Level Objective (SLO). It proactively identifies critical microservices and performs dynamic reprovisioning to mitigate potential SLO violations. Testbed experiments show that AutoMan can save CPU and memory resources by up to 49.6% and 29.1% on average while ensuring the same end-to-end tail latency objective.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2023)
Article
Chemistry, Analytical
Zaakki Ahamed, Maher Khemakhem, Fathy Eassa, Fawaz Alsolami, Abdullah Basuhail, Kamal Jambi
Summary: In this paper, a novel solution called FEDQWP is proposed, which leverages deep Q-learning to predict federated cloud workload. The solution comprehensively addresses the issues of VM placement, energy efficiency, and SLA adherence, and the experimental results demonstrate its superiority in optimizing performance.
Article
Automation & Control Systems
Jatin Bedi, Yashwant Singh Patel
Summary: In this paper, a lightweight storage workload time series prediction method named 'STOWP' is proposed, which integrates the Neural Basis Expansion Analysis (N-BEATS) deep model with windowing strategy to improve job scheduling and load balancing. Experimental results show that 'STOWP' has achieved significant improvements in RMSE and MAE compared to existing techniques.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2022)
Article
Computer Science, Information Systems
Ali Asghari, Mohammad Karim Sohrabi
Summary: The proposed method, MDQ-CR, combines coral reefs optimization algorithm and multi-agent deep Q-network to reduce the energy consumption of data centers and cloud resources. By utilizing techniques like dynamic voltage and frequency scaling for processors, the method shows higher energy savings compared to other methods in empirical experiments.
CLUSTER COMPUTING-THE JOURNAL OF NETWORKS SOFTWARE TOOLS AND APPLICATIONS
(2022)
Article
Computer Science, Information Systems
Jiechao Gao, Haoyu Wang, Haiying Shen
Summary: Large-scale cloud data centers often face high failure rates due to hardware and software failures, which can greatly reduce service reliability and require significant resources for recovery. Predicting task and job failures with high accuracy is crucial to avoid wastage. This article proposes a failure prediction algorithm based on multi-layer Bi-LSTM, which outperforms other methods with 93% accuracy for task failure and 87% accuracy for job failures.
IEEE TRANSACTIONS ON SERVICES COMPUTING
(2022)
Article
Computer Science, Information Systems
Shuiguang Deng, Yishan Chen, Gong Chen, Shouling Ji, Jianwei Yin, Albert Y. Zomaya
Summary: This paper presents a proactive application deployment system consisting of three modules (incentive, profit, and latency) that optimizes application deployment based on a fully distributed edge network architecture. The SELL algorithm in the incentive module allows edge servers to compete for deployment rights in a two-stage Stackelberg game and receive payment for their efforts. The other two modules recursively adjust service prices and deployment intentions based on their own profits. Simulations demonstrate that the SELL algorithm can help application providers find suitable edge servers for deployment while maximizing profits for both parties with low latency.
IEEE TRANSACTIONS ON MOBILE COMPUTING
(2023)
Article
Computer Science, Theory & Methods
Hailiang Zhao, Shuiguang Deng, Feiyi Chen, Jianwei Yin, Schahram Dustdar, Albert Y. Zomaya
Summary: This article discusses the scheduling of multi-server jobs online and proposes the Esdp algorithm to deal with the unknown actual processing speeds. By learning the distribution of processing speed fluctuations, the Esdp algorithm maximizes the cumulative overall utility and has polynomial complexity and logarithmic regret.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
(2023)
Article
Computer Science, Information Systems
Tianming Zhao, Wei Li, Boyu Qin, Ling Wang, Albert Y. Zomaya
Summary: This research focuses on the coordination problem of maximizing Pulsed Power Load (PPL) utility while maintaining normal loads in isolated power systems. Two scenarios, fixed and general normal loads, are considered and dynamic programming algorithms are developed to optimize the coordination schedule. Experimental evaluation confirms the practicality of the solutions.
ACM TRANSACTIONS ON MODELING AND PERFORMANCE EVALUATION OF COMPUTING SYSTEMS
(2023)
Article
Computer Science, Information Systems
Firas Al-Doghman, Nour Moustafa, Ibrahim Khalil, Nasrin Sohrabi, Zahir Tari, Albert Y. Zomaya
Summary: The paradigm of edge computing has created new possibilities for the Internet of Things (IoT), expanding cloud services to the network edge. This allows for the design of distributed architectures and the enhancement of decision-making applications. Edge computing faces challenges related to security and privacy, but advancements in artificial intelligence and machine learning provide opportunities for precise models and intelligent applications at the network edge. This study presents a comprehensive survey on securing edge computing-based AI microservices, highlighting key requirements and proposing a secure edge computing framework.
IEEE TRANSACTIONS ON SERVICES COMPUTING
(2023)
Article
Engineering, Electrical & Electronic
Yongchao Zhang, Jia Hu, Geyong Min
Summary: The paper proposes a digital twin-driven intelligent task offloading framework for collaborative mobile edge computing (MEC). By mapping the MEC system into a virtual space using digital twin and optimizing task offloading decisions with deep reinforcement learning, the proposed framework effectively adapts to dynamic environments and significantly improves the MEC system's income.
IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS
(2023)
Article
Computer Science, Hardware & Architecture
Rui Jin, Jia Hu, Geyong Min, Jed Mills
Summary: Federated Learning is a privacy-preserving distributed machine learning method, but it faces challenges like malicious clients and communication overhead. To address these challenges, a lightweight Blockchain-Empowered Federated Learning system is proposed, which integrates secure and efficient training scheme, consensus mechanism, and scalable blockchain architecture.
IEEE TRANSACTIONS ON COMPUTERS
(2023)
Article
Automation & Control Systems
Xiaoding Wang, Jia Hu, Hui Lin, Wenxin Liu, Hyeonjoon Moon, Md. Jalil Piran
Summary: The integration of IoT and the medical industry has led to the emergence of IoMT. In IoMT, physicians analyze patient data collected through mobile devices with the assistance of AI systems. However, traditional AI technologies may compromise patient privacy. To address this issue, we propose a privacy-enhanced disease diagnosis mechanism using federated learning in IoMT.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Engineering, Civil
Cheng Qiao, Jing Qiu, Zhiyuan Tan, Geyong Min, Albert Y. Zomaya, Zhihong Tian
Summary: This paper studies the problem of performance evaluation in IoV and proposes a general approach to measure the performance of individual agents by exploring the common knowledge and correlation between different agents. Experimental results show that our evaluation scheme is efficient in these settings.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Computer Science, Theory & Methods
Yuchen Li, Weifa Liang, Jing Li, Xiuzhen Cheng, Dongxiao Yu, Albert Y. Zomaya, Song Guo
Summary: The rise of deep learning brings new vitality to the future of intelligent IoT, and the emergence of edge intelligence enables real-time DNN inference services for mobile users. To ensure efficient and secure DNN model training in edge computing, federated learning is proposed as an ideal learning paradigm. This article focuses on energy-aware DNN model training in edge computing and proposes an algorithm to optimize the global loss of the training model while considering bandwidth and energy constraints.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
(2023)
Article
Computer Science, Theory & Methods
Jed Mills, Jia Hu, Geyong Min
Summary: In Federated Learning, a new approach is proposed to gradually reduce the number of steps K of Stochastic Gradient Descent (SGD) performed on clients per round during training. This can improve the performance of the FL model while reducing the training time and computational cost. Thorough experiments on benchmark FL datasets demonstrate the real-world benefits of this approach in terms of convergence time, computational cost, and generalization performance.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
(2023)
Article
Computer Science, Information Systems
Marwa Keshk, Nickolaos Koroniotis, Nam Pham, Nour Moustafa, Benjamin Turnbull, Albert Y. Zomaya
Summary: Although XAI has gained significant interest, its implementation in cyber security applications needs further investigation. This paper proposes a novel explainable intrusion detection framework for IoT networks, using a LSTM model and a novel SPIP framework for training and evaluating the model. The SPIP framework achieves high detection accuracy, processing time, and interpretability of data features and model outputs.
INFORMATION SCIENCES
(2023)
Article
Computer Science, Information Systems
Xing Chen, Zewei Yao, Zheyi Chen, Geyong Min, Xianghan Zheng, Chunming Rong
Summary: Mobile edge computing (MEC) reduces latency and energy consumption of mobile applications by offloading tasks to nearby edges. This study proposes a novel two-stage decision-making method for load balancing in multiedge collaboration. It combines centralized decision-making with global information and decentralized decision-making with local information to achieve optimal load balancing.
IEEE INTERNET OF THINGS JOURNAL
(2023)
Article
Computer Science, Information Systems
Zheyi Chen, Pu Tian, Weixian Liao, Xuhui Chen, Guobin Xu, Wei Yu
Summary: This article introduces a knowledge transfer-based federated learning framework under a resource-limited distributed system to address the challenges of federated learning in IoT systems. The proposed approach uses knowledge distillation to improve efficiency and outperform other schemes.
IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTING
(2023)
Article
Computer Science, Information Systems
Jin Wang, Jia Hu, Geyong Min, Qiang Ni, Tarek El-Ghazawi
Summary: The research proposes a novel learning-driven method with a user-centric approach to address service migration in dynamic MEC environments. By modeling the problem as a Partially Observable Markov Decision Process (POMDP), a new encoder network combining LSTM and an embedding matrix is designed for effective information extraction, and a tailored off-policy actor-critic algorithm is proposed for efficient training.
IEEE TRANSACTIONS ON MOBILE COMPUTING
(2023)
Article
Engineering, Civil
Xiaoding Wang, Sahil Garg, Hui Lin, Georges Kaddoum, Jia Hu, Mohammad Mehedi Hassan
Summary: This paper proposes a hierarchical trust evaluation strategy based on heterogeneous blockchain, utilizing federated deep learning technology for Intelligent Transportation Systems (ITS) security.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)