Developing a hybrid deep learning model for workload prediction in virtualized data centers

The relevance of this research stems from the need to proactively manage the workload of data centers based on virtual machine technologies and application containerization. This task requires analyzing historical workload data consolidated as time series for computing resources used over a given period of time, such as CPU and RAM pool utilization and stored by the resource monitoring subsystem of the data center administration service. Therefore, this article aims to explore machine learning methods and technologies that support time series forecasting. The article summarizes the features of machine learning models based on statistical approaches and deep learning principles. It examines the structural and functional components of neural network variants specialized in analyzing time series dependencies and solving forecasting problems. The proposed solution is a hybrid deep learning system based on the sequential application of cascades of one-dimensional convolutional neural networks and bidirectional long short-term memory networks. Approaches to the selection of their structural and parametric characteristics are proposed. The results of a comparative experimental evaluation of the proposed solution with the implementation of a workload prediction system, based on statistical prediction methods are presented.

1. Shen L., Qian Sh., Zhai T., Li L., Li Zh. Research on Cloud Computing High-Density Data Center Infrastructure and Environment Matching Technology. In: 2020 2nd International Conference on Computer Science Communication and Network Security (CSCNS2020): MATEC Web of Conferences: Volume 336, 22–23 December 2020, Sanya, China. EDP Sciences; 2012. https://doi.org/10.1051/matecconf/202133602028

2. Uddin M., Rahman A.A., Shah A., Memon J. Virtualization Implementation Approach for Data Centers to Maximize Performance. Asian Journal of Scientific Research. 2012;5(2):45–57. https://doi.org/10.3923/ajsr.2012.45.57

3. Cox-Fuenzalida L.-E. Effect of Workload History on Task Performance. Human Factors: The Journal of the Human Factors and Ergonomics Society. 2007;49(2):277–291. https://doi.org/10.1518/001872007X312496

4. Tran V.G., Debusschere V., Bacha S. Hourly Server Workload Forecasting up to 168 Hours Ahead Using Seasonal ARIMA Model. In: 2012 IEEE International Conference on Industrial Technology, 19–21 March 2012, Athens, Greece. IEEE; 2012. P. 1127–1131. https://doi.org/10.1109/ICIT.2012.6210091

5. Sun Q., Tan Zh., Zhou X. Workload Prediction of Cloud Computing Based on SVM and BP Neural Networks. Journal of Intelligent & Fuzzy Systems: Applications in Engineering and Technology. 2020;39(3):2861–2867. https://doi.org/10.3233/JIFS-191266

6. Nguyen H.M., Kalra G., Kim D. Host Load Prediction in Cloud Computing Using Long Short-Term Memory Encoder-Decoder. The Journal of Supercomputing. 2019;75(11):7592–7605. https://doi.org/10.1007/s11227-019-02967-7

7. Pashshoev B., Petrusevich D.A. Neural Network Analysis in Time Series Forecasting. Russian Technological Journal. 2024;12(4):106–116. https://doi.org/10.32362/2500-316X-2024-12-4-106-116

8. Mitiche I., Nesbitt A., Conner S., Boreham Ph., Morison G. 1D-CNN Based Real-Time Fault Detection System for Power Asset Diagnostics. IET Generation, Transmission & Distribution. 2020;14(24):5766–5773. https://doi.org/10.1049/iet-gtd.2020.0773

9. Wibawa A.P., Fadhilla A.F., Paramarta A.Kh.I., et al. Bidirectional Long Short-Term Memory (Bi-LSTM) Hourly Energy Forecasting. In: International Conference on Computer Science Electronics and Information (ICCSEI 2023): E3S Web of Conferences, Volume 501, 12–13 December 2023, Yogyakarta, Indonesia. EDP Sciences; 2024. https://doi.org/10.1051/e3sconf/202450101023

10. Ban Y., Zhang D., He Q., Shen Q. APSO-CNN-SE: An Adaptive Convolutional Neural Network Approach for IoT Intrusion Detection. Computers, Materials and Continua. 2024;81(1):567–601. https://doi.org/10.32604/cmc.2024.055007

11. Rasheduzzaman M., Islam A., Rahman R.M. Workload Prediction on Google Cluster Trace. International Journal of Grid and High Performance Computing. 2014;6(3):34–52. https://doi.org/10.4018/ijghpc.2014070103

12. Almalchy M.T., Ciobanu V., Popescu N. Noise Removal from ECG Signal Based on Filtering Techniques. In: 2019 22nd International Conference on Control Systems and Computer Science (CSCS), 28–30 May 2019, Bucharest, Romania. IEEE; 2019. P. 176–181. https://doi.org/10.1109/CSCS.2019.00037