VTGAN: hybrid generative adversarial networks for cloud workload prediction

Journal of Cloud Computing Advances Systems and Applications, Год журнала: 2023, Номер 12(1)

Опубликована: Июнь 26, 2023

Abstract Efficient resource management approaches have become a fundamental challenge for distributed systems, especially dynamic environment systems such as cloud computing data centers. These aim at load-balancing or minimizing power consumption. Due to the highly nature of workloads, traditional time series and machine learning models fail achieve accurate predictions. In this paper, we propose novel hybrid VTGAN models. Our proposed not only predicting future workloads but also workload trend (i.e., upward downward direction workload). Trend classification could be less complex during decision-making process in approaches. Also, study effect changing sliding window size number prediction steps. addition, investigate impact enhancing features used training using technical indicators, Fourier transforms, wavelet transforms. We validate our real dataset. results show that outperform deep models, LSTM/GRU CNN-LSTM/GRU, concerning classification. model records an accuracy ranging from $$95.4\%$$ 95.4 % $$96.6\%$$ 96.6 .

Язык: Английский

---

Robustness of Workload Forecasting Models in Cloud Data Centers: A White-Box Adversarial Attack Perspective

IEEE Access, Год журнала: 2024, Номер 12, С. 55248 - 55263

Опубликована: Янв. 1, 2024

Cloud computing has become the cornerstone of modern technology, propelling industries to unprecedented heights with its remarkable and recent advances. However, fundamental challenge for cloud service providers is real-time workload prediction management optimal resource allocation. workloads are characterized by their heterogeneous, unpredictable, fluctuating nature, making this task even more challenging. As a result achievements deep learning (DL) algorithms across diverse fields, scholars have begun embrace approach addressing such challenges. It defacto standard prediction. Unfortunately, DL been widely recognized vulnerability adversarial examples, which poses significant DL-based forecasting models. In study, we utilize established white-box attack generation methods from field computer vision construct examples four cutting-edge regression models, including Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), Gated Unit (GRU), 1D Convolutional (1D-CNN) attention-based We evaluate our study three benchmark datasets: Google trace, Alibaba Bitbrain. The findings analysis unequivocally indicate that models highly vulnerable attacks. To best knowledge, first conduct systematic research exploring in data center, highlighting inherent hazards both security cost-effectiveness centers. By raising awareness these vulnerabilities, advocate urgent development robust defensive mechanisms enhance constantly evolving technical landscape.

Язык: Английский

---

Enhancing Long-Term Cloud Workload Forecasting Framework: Anomaly Handling and Ensemble Learning in Multivariate Time Series

IEEE Transactions on Cloud Computing, Год журнала: 2024, Номер 12(2), С. 789 - 799

Опубликована: Апрель 1, 2024

Forecasting workloads and responding promptly with resource scaling migration is critical to optimizing operations enhancing management in cloud environments. However, the diverse dynamic nature of devices within environments complicates workload forecasting. These challenges often lead service level agreement violations or inefficient usage. Hence, this paper proposes an Enhanced Long-Term Cloud Workload (E-LCWF) framework designed specifically for efficient these heterogeneous The E-LCWF processes individual as multivariate time series enhances model performance through anomaly detection handling. Additionally, employs error-based ensemble approach, using transformer-based models Time Series (LTSF) linear models, each which has demonstrated exceptional LTSF. Experimental results obtained virtual machine data from real-world information systems manufacturing execution show that outperforms state-of-the-art forecasting accuracy.

Язык: Английский

---

Application-Oriented Cloud Workload Prediction: A Survey and New Perspectives

Tsinghua Science & Technology, Год журнала: 2024, Номер 30(1), С. 34 - 54

Опубликована: Сен. 11, 2024

Язык: Английский

---

Automatic data featurization for enhanced proactive service auto-scaling: Boosting forecasting accuracy and mitigating oscillation

Journal of King Saud University - Computer and Information Sciences, Год журнала: 2024, Номер 36(2), С. 101924 - 101924

Опубликована: Янв. 21, 2024

Edge computing has gained widespread adoption for time-sensitive applications by offloading a portion of IoT system workloads from the cloud to edge nodes. However, limited resources devices hinder service deployment, making auto-scaling crucial improving resource utilization in response dynamic workloads. Recent solutions aim make proactive predicting future and overcoming limitations reactive approaches. These often rely on time-series data analysis machine learning techniques, especially Long Short-Term Memory (LSTM), thanks its accuracy prediction speed. existing suffer oscillation issues, even when using cooling-down strategy. Consequently, efficiency depends model degree scaling actions. This paper proposes novel approach improve deal with issues. Our involves an automatic featurization phase that extracts features workload data, prediction's accuracy. extracted also serve as grid controlling generated experimental results demonstrate effectiveness our accuracy, mitigating phenomena, enhancing overall performance.

Язык: Английский

---

Deep Learning-based Workload Prediction in Cloud Computing to Enhance the Performance

Опубликована: Май 26, 2023

Язык: Английский

---

A feature extraction and time warping based neural expansion architecture for cloud resource usage forecasting

Cluster Computing, Год журнала: 2024, Номер 27(4), С. 4963 - 4982

Опубликована: Янв. 10, 2024

Язык: Английский

---

Hybrid deep learning and evolutionary algorithms for accurate cloud workload prediction

Computing, Год журнала: 2024, Номер 106(12), С. 3905 - 3944

Опубликована: Авг. 25, 2024

Язык: Английский

---

ADAPT: Attention-Driven Domain Adaptation for Inter-cluster Workload Forecasting in Cloud Data Centers

Lecture notes in computer science, Год журнала: 2024, Номер unknown, С. 56 - 68

Опубликована: Ноя. 15, 2024

Язык: Английский

---

Deep Learning Neural Networks in the Cloud

International Journal of Advanced engineering Management and Science, Год журнала: 2023, Номер 9(10), С. 09 - 26

Опубликована: Янв. 1, 2023

Deep Neural Networks (DNNs) are currently used in a wide range of critical real-world applications as machine learning technology. Due to the high number parameters that make up DNNs, and prediction tasks require millions floating-point operations (FLOPs). Implementing DNNs into cloud computing system with centralized servers data storage sub-systems equipped high-speed high-performance capabilities is more effective strategy. This research presents an updated analysis most recent computing. It highlights necessity while presenting debating numerous DNN complexity issues related various architectures. Additionally, it goes their intricacies offers thorough several platforms for deployment. examines already running on highlight advantages using DNNs. The study difficulties associated implementing systems provides suggestions improving both current future deployments.

Язык: Английский

---