Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CloudFormer: An Attention-based Performance Prediction for Public Clouds with Unknown Workload

Amirhossein Shahbazinia, Darong Huang, Luis Costero, David Atienza

Abstract

Cloud platforms are increasingly relied upon to host diverse, resource-intensive workloads due to their scalability, flexibility, and cost-efficiency. In multi-tenant cloud environments, virtual machines are consolidated on shared physical servers to improve resource utilization. While virtualization guarantees resource partitioning for CPU, memory, and storage, it cannot ensure performance isolation. Competition for shared resources such as last-level cache, memory bandwidth, and network interfaces often leads to severe performance degradation. Existing management techniques, including VM scheduling and resource provisioning, require accurate performance prediction to mitigate interference. However, this remains challenging in public clouds due to the black-box nature of VMs and the highly dynamic nature of workloads. To address these limitations, we propose CloudFormer, a dual-branch Transformer-based model designed to predict VM performance degradation in black-box environments. CloudFormer jointly models temporal dynamics and system-level interactions, leveraging 206 system metrics at one-second resolution across both static and dynamic scenarios. This design enables the model to capture transient interference effects and adapt to varying workload conditions without scenario-specific tuning. Complementing the methodology, we provide a fine-grained dataset that significantly expands the temporal resolution and metric diversity compared to existing benchmarks. Experimental results demonstrate that CloudFormer consistently outperforms state-of-the-art baselines across multiple evaluation metrics, achieving robust generalization across diverse and previously unseen workloads. Notably, CloudFormer attains a mean absolute error (MAE) of just 7.8%, representing a substantial improvement in predictive accuracy and outperforming existing methods at least by 28%.

CloudFormer: An Attention-based Performance Prediction for Public Clouds with Unknown Workload

Abstract

Cloud platforms are increasingly relied upon to host diverse, resource-intensive workloads due to their scalability, flexibility, and cost-efficiency. In multi-tenant cloud environments, virtual machines are consolidated on shared physical servers to improve resource utilization. While virtualization guarantees resource partitioning for CPU, memory, and storage, it cannot ensure performance isolation. Competition for shared resources such as last-level cache, memory bandwidth, and network interfaces often leads to severe performance degradation. Existing management techniques, including VM scheduling and resource provisioning, require accurate performance prediction to mitigate interference. However, this remains challenging in public clouds due to the black-box nature of VMs and the highly dynamic nature of workloads. To address these limitations, we propose CloudFormer, a dual-branch Transformer-based model designed to predict VM performance degradation in black-box environments. CloudFormer jointly models temporal dynamics and system-level interactions, leveraging 206 system metrics at one-second resolution across both static and dynamic scenarios. This design enables the model to capture transient interference effects and adapt to varying workload conditions without scenario-specific tuning. Complementing the methodology, we provide a fine-grained dataset that significantly expands the temporal resolution and metric diversity compared to existing benchmarks. Experimental results demonstrate that CloudFormer consistently outperforms state-of-the-art baselines across multiple evaluation metrics, achieving robust generalization across diverse and previously unseen workloads. Notably, CloudFormer attains a mean absolute error (MAE) of just 7.8%, representing a substantial improvement in predictive accuracy and outperforming existing methods at least by 28%.

Paper Structure

This paper contains 33 sections, 3 equations, 5 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. VM Performance Challenges and Mitigation
  4. Performance Prediction Methods
  5. Performance Monitoring and Prediction Datasets
  6. Problem Definition
  7. CloudPerfTracefoot:ds: A Comprehensive Dataset for Performance Analysis
  8. Testbed Server Setup
  9. Profiling Features
  10. Application Scenarios
  11. Data
  12. CloudFormer: A Unified Approach to Black-Box VM Performance Prediction
  13. Input Representation
  14. Temporal Branch
  15. System Branch
  16. ...and 18 more sections

Figures (5)

  • Figure 1: General overview of the performance degradation prediction problem in multi-tenant cloud environments. Multiple VMs compete for shared resources, causing interference and variable execution times. Due to the black-box nature of public clouds, only system-level traces observable from the host can be used for modeling and prediction.
  • Figure 2: For applications with dynamic workloads, four distinct scenarios are considered: (a) static workload, (b) monotonic workload, (c) periodic workload, and (d) random workload.
  • Figure 3: Overview of CloudFormer architecture illustrating dual branches for temporal and system-level modeling.
  • Figure 4: Stacked bar plot illustrating the distribution of prediction errors across different models, showing the percentage of samples falling within specific error segments.
  • Figure 5: Heatmaps showing the mean absolute error (MAE $\pm$ STD) of different models across six random seeds and four test applications (two static-only and two dynamic scenarios). For better visual contrast, the color bar is capped at 20% MAE.