Umeå University's Workload Buoyancy detects resource bottlenecks 19.3% better

Researchers from Umeå University and Unaffiliated contributors have proposed Workload Buoyancy, a breakthrough metric for identifying performance bottlenecks in complex multi-tenant computing environments. Published on arXiv, this work addresses a critical gap in modern cloud infrastructure where traditional heuristics—like CPU utilization—fail to capture the nuanced performance dynamics caused by resource contention and 'noisy neighbor' effects.

The Buoyancy metric integrates both application-level and system-level data to provide a holistic view of performance across shared resources like memory, network bandwidth, and storage I/O. In evaluations with representative workloads, Buoyancy outperformed conventional heuristics by 19.3% in accurately identifying bottlenecks, offering a more intuitive and generalizable framework for resource-aware orchestration. This improved observability enables cloud providers and DevOps teams to make more informed scheduling and optimization decisions, reducing unexpected degradation and improving application reliability in heterogeneous environments.