Performance of Zero-Shot Time Series Foundation Models on Cloud Data

TL;DR

This paper evaluates the effectiveness of zero-shot time series foundation models on cloud data, revealing they often underperform simple baselines and exhibit unpredictable behavior.

Contribution

It provides the first systematic analysis of FMs on cloud data, highlighting their limitations and failure modes in this domain.

Findings

FMs are outperformed by simple linear models on cloud data.

Many FMs produce erratic and meaningless forecasts.

FMs generally fail to model cloud data effectively.

Abstract

Time series foundation models (FMs) have emerged as a popular paradigm for zero-shot multi-domain forecasting. FMs are trained on numerous diverse datasets and claim to be effective forecasters across multiple different time series domains, including cloud data. In this work we investigate this claim, exploring the effectiveness of FMs on cloud data. We demonstrate that many well-known FMs fail to generate meaningful or accurate zero-shot forecasts in this setting. We support this claim empirically, showing that FMs are outperformed consistently by simple linear baselines. We also illustrate a number of interesting pathologies, including instances where FMs suddenly output seemingly erratic, random-looking forecasts. Our results suggest a widespread failure of FMs to model cloud data.