Dynamic Splitting of Diffusion Models for Multivariate Time Series Anomaly Detection in a JointCloud Environment

Speeding up multivariate time series anomaly detection models in the JointCloud environment is a challenging task. Typically, anomaly detection models with high computational requirements are offloaded to cross-cloud nodes with abundant computational resources, which can accelerate the inference procedure. However, this cross-cloud task offloading is hampered by the instability of network conditions in the JointCloud environment. Meanwhile, compressing the anomaly detection model to a size suitable for being deployed onto a cloud node close to the application scenario may lead to dramatic performance loss. To overcome these challenges, dynamic splitting anomaly detection models emerge as a promising solution by utilizing the computational power of various cross-cloud nodes to reduce the computational cost while maintaining their performance. To this end, in this paper, we propose a new dynamic splitting method tailored for time series anomaly detection models with self-adaptation to dynamic fluctuations in network bandwidth in the JointCloud environment. Moreover, both single-objective and multi-objective optimizations are formulated with solutions provided in detail. By conducting comprehensive experiments on 3 public datasets in a JointCloud environment, our proposed dynamic splitting method outperforms the baselines in terms of achieving a better balance between inference speed and energy consumption.