ADARM: An Application-Driven Adaptive Resource Management Framework for Data Centers
Min Luo, Li Li, Wu Chou · 2017
ADARM is a framework for application-driven and adaptive resource allocation to enable large scale service-oriented cloud DCs. It builds and utilizes real-time application profiles and models to understand application requirements and performance characteristics. In addition, service-orientation enables ADARM to use uniform and flexible REST services to abstract, organize, monitor and control the fine-grained cloud resources, and it separates the resource control logic from the datacenter infrastructure. The combination of application-driven and service-orientation enables ADARM to incorporate various adaptive optimization techniques to optimize resource allocations for large scale distributed DCs. This paper discuses the design rationales behind ADARM, and describes its architecture and main benefits. Prototypical validations on several core components in ADARM are presented to demonstrate the potential business value and the feasibility of such a system framework.