Towards evaluating holistic runtime state based on change vector analysis
Huihong He, Qian Liu, Lingling Li, Zhao Li, Zhiyi Ma, Yong Wang, Dongjin Fan · 2017
Continuous evolution of virtualization-related technologies have benefit greatly flexibility and reliability of software system. However, although virtualization removes binding between application and server underneath, it complicates the regular runtime hierarchy into an intricate structure, where cross-layer monitoring and holistic evaluation becomes challenging. Currently lots of monitoring practices take hierarchical pattern and most of research work hasn't considered intricate structure, so we in this paper make a preliminary exploration and attempt on cross-layer evaluation. An approach is proposed to evaluate host runtime state based on change vector analysis method. Our approach learns to quantitate the interaction among software, container and server layers from large cross-layer monitoring data, and evaluates host state based on simple commonsense rules and quantitative observations. So far, our approach has demonstrated effectiveness in evaluating hosts in IaaS cloud.