Multiscale Representation Enhanced Temporal Flow Fusion Model for Long-Term Workload Forecasting

TL;DR

The paper addresses long-term workload forecasting in cloud systems, which is challenged by non-stationary, nonlinear time series and long-range dependencies. It introduces a two-stage approach: first, self-supervised multiscale time series representation learning in time and frequency domains; second, a Temporal Flow Fusion Model that combines long-term representations with near-term observations via FusionAttention and a conditional normalizing flow for probabilistic forecasting. Key contributions include (1) a novel multiscale representation method with time- and frequency-domain contrastive losses, (2) a fusion-based predictor that integrates long-term history and near-term dynamics, and (3) extensive evaluations across 9 benchmarks showing state-of-the-art performance, plus real-world deployment in Alipay yielding substantial resource savings. This work advances accurate, scalable, and probabilistic long-term workload forecasting to improve autoscaling and cloud resource management in practical settings.

Abstract

Accurate workload forecasting is critical for efficient resource management in cloud computing systems, enabling effective scheduling and autoscaling. Despite recent advances with transformer-based forecasting models, challenges remain due to the non-stationary, nonlinear characteristics of workload time series and the long-term dependencies. In particular, inconsistent performance between long-term history and near-term forecasts hinders long-range predictions. This paper proposes a novel framework leveraging self-supervised multiscale representation learning to capture both long-term and near-term workload patterns. The long-term history is encoded through multiscale representations while the near-term observations are modeled via temporal flow fusion. These representations of different scales are fused using an attention mechanism and characterized with normalizing flows to handle non-Gaussian/non-linear distributions of time series. Extensive experiments on 9 benchmarks demonstrate superiority over existing methods.

Figures (6)

• Figure 1: Framework Architecture.
• Figure 2: Multiscale Time Series Representation
• Figure 3: Temporal Flow Fusion Model Architecture.
• Figure 4: Visualization of TS representation of different windows in Workload TS.
• Figure 5: Visualization the clusters of Workload TS representation.