Benchmark RadarCloudCons
CloudCons is an end-to-end benchmark for evaluating forecasting models in cloud resource consolidation. It encompasses datasets from Huawei Cloud, Microsoft Azure, and Google Borg with diverse workload patterns, and evaluates statistical, deep learning, and time series foundation models. The evaluation includes resource efficiency and service reliability metrics under a forecast-then-optimize paradigm.
- Released
- 2026-06-11
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Existing benchmarks focus only on prediction error, leaving the downstream decision utility of forecasting models unverified. CloudCons addresses this gap by assessing practical value in cloud resource consolidation, offering insights into balancing resource efficiency and service reliability, aiding deployment decisions.
Motivation
Driven by conservative over-provisioning to guarantee service reliability, resource utilization in cloud data centers remains at low levels.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.