The objective of this project is to estimate from measurements the capacity region for a cloud RAN and to demonstrate the usefulness of this concept for resource allocation and anomaly detection. The capacity region characterizes the load patterns that a given infrastructure can support for a specific service and SLA (Service-Level Agreement). In the project, a model of the capacity region will be learned from infrastructure statistics and service monitoring data. The model will be recomputed after infrastructure changes or reconfigurations. The validity of the concept will be evaluated through scenarios with changing external load, and its applicability will be demonstrated through use cases involving proactive resource provisioning or bottleneck avoidance. The concept of a capacity region has been investigated before, mainly in the context of analytical modeling of wireless networks. The novel aspect of this work, from a research perspective, lies in the real-time estimation of the capacity through measurements and an adaptive learning process.
Scope: Real Time Analytics
Main KTH Staff Involved: Rolf Stadler
Other Partners: Ericsson