Skip to content
DraconiticResearch
Search Sources API
203,952 studies31 sourcesUpdated 47 min ago
PreprintCC BY 4.0Zenodo

Multi-Scale Temporal Modeling with Volatility-Aware Channel Attention for Multi-Step Cloud Workload Forecasting

  • Liang Zheng1
  • Zhongning Guo2
  • Ming Wu3
  1. 1School of Computer Science, Shanghai Jiao Tong University, Shanghai 200240, China
  2. 2College of Mechanical and Electrical Engineering, Guangdong University of Technology, Guangzhou 510006, China
  3. 3Department of Computer Science, KU Leuven, 3001 Leuven, Belgium
Open at sourcePDFDOICite

Abstract

Cloud platforms require forecasts of CPU, memory and network demand several steps ahead because capacity adjustments take time. Workload levels and variability drift, short bursts coexist with slower changes, and the usefulness of each resource channel varies across windows. This preprint describes reversible window normalization, attention conditioned on window mean and variance, and three paths for trend, convolution–recurrent dynamics and multi-scale residuals. The residual path uses kernels of sizes 3 and 7; sigmoid gates regulate its contribution and the trend correction. The supplied evaluation summary describes an Alibaba cluster trace, a 96-step history and horizons of 1, 5, 10 and 30 steps. It reports the lowest MAE among five methods at every horizon, including a 4.1% reduction relative to C-LSTM at one step and reductions of 7.3%–8.3% at longer horizons. It reports the lowest MSE from five steps onward, while two recurrent baselines perform slightly better at one step. Component removal, severe input corruption and post-spike transitions degrade performance. The manuscript formalizes one compatible architecture and an evaluation protocol. These are source-reported findings: absolute errors, uncertainty estimates and executable experiment artifacts were not supplied and the results have not been independently reproduced. The PDF and editable LaTeX source package are included.

Related

  1. Improved Dung Beetle Optimizer–Tuned Convolutional–Recurrent Temporal Self-Attention Network for Multi-Horizon Traffic Speed Forecasting
    Jiwu Tang, Linchao An, Haifeng Liu +1
    Sensors (Basel, Switzerland)Peer-reviewedCC BYFull textEurope PMC
  2. STCFormer: Adaptive Spatio-Temporal Modeling with Dynamic Cluster Transformer for Station-based Weather Forecasting
    Rongwen Li, Haixin Xie, Mingyang Wang +5
    arXivPreprintCC BY 4.0Full textStatisticsarXiv
  3. Short-term metro passenger flow forecasting via decoupled spatio-temporal modeling with a diffusion-convergence graph and Swin-Enhanced LSTM
    Yixin Feng, Shangbing Gao, Hao Wang +5
    PLOS OnePeer-reviewedCC BY 4.0Full textCell biologyNeurosciencePLOS
  4. Graph attention spatio-temporal graph neural network based human–light interaction modeling for indoor lighting environments
    Ting Gao, Lingling Sun
    PLOS OnePeer-reviewedCC BY 4.0Full textNeurosciencePLOS
  5. Uncertainty-aware hybrid autoscaling with bias-corrected deep forecasting for cost-efficient capacity-shortfall control across cloud workload traces
    Mahdee Jodayree, Puya Shaykholeslami, Arman Kavoosi Ghafi +2
    Scientific ReportsPeer-reviewedCC BY-NC-ND 4.0Full textCrossref

Cite

BibTeXRIS
BibTeX
@misc{zheng2026multiscale,
  title = {{Multi-Scale Temporal Modeling with Volatility-Aware Channel Attention for Multi-Step Cloud Workload Forecasting}},
  author = {Liang Zheng and Zhongning Guo and Ming Wu},
  howpublished = {Zenodo},
  year = {2026},
  month = oct,
  doi = {10.5281/zenodo.23580729},
  url = {https://zenodo.org/records/23580729},
  note = {CC BY 4.0},
}
RIS
TY  - UNPB
TI  - Multi-Scale Temporal Modeling with Volatility-Aware Channel Attention for Multi-Step Cloud Workload Forecasting
AU  - Liang Zheng
AU  - Zhongning Guo
AU  - Ming Wu
PY  - 2026
DA  - 2026/10/22
PB  - Zenodo
DO  - 10.5281/zenodo.23580729
UR  - https://zenodo.org/records/23580729
AB  - Cloud platforms require forecasts of CPU, memory and network demand several steps ahead because capacity adjustments take time. Workload levels and variability drift, short bursts coexist with slower changes, and the usefulness of each resource channel varies across windows. This preprint describes reversible window normalization, attention conditioned on window mean and variance, and three paths for trend, convolution–recurrent dynamics and multi-scale residuals. The residual path uses kernels of sizes 3 and 7; sigmoid gates regulate its contribution and the trend correction. The supplied evaluation summary describes an Alibaba cluster trace, a 96-step history and horizons of 1, 5, 10 and 30 steps. It reports the lowest MAE among five methods at every horizon, including a 4.1% reduction relative to C-LSTM at one step and reductions of 7.3%–8.3% at longer horizons. It reports the lowest MSE from five steps onward, while two recurrent baselines perform slightly better at one step. Component removal, severe input corruption and post-spike transitions degrade performance. The manuscript formalizes one compatible architecture and an evaluation protocol. These are source-reported findings: absolute errors, uncertainty estimates and executable experiment artifacts were not supplied and the results have not been independently reproduced. The PDF and editable LaTeX source package are included.
ER  -