Abstract

Despite modern advances in automation and managed services, many end users of cloud services remain concerned about the lack of visibility into their operational environments. The underlying principles of existing approaches employed to aid users gain visibility into their runtimes do not apply to today's dynamic cloud environment where virtual machines and containers operate as processes of the cloud operating system. We present near field monitoring (NFM), a cloud-native framework for monitoring cloud systems and providing operational analytics services. With NFM, we employ cloud, virtualization, and containerization abstractions to provide extensive visibility into running entities in the cloud, in a touchless manner, i.e., without modifying, instrumenting, or accessing within the end-user context. Operating outside the context of the target systems enables always-on monitoring independent of their health. Using an NFM implementation on OpenStack, we demonstrate the capabilities of NFM, as well as its monitoring accuracy and efficiency. NFM is practical and general, supporting more than 1,000 different system distributions, allowing instantaneous monitoring as soon as a guest system becomes hosted on the cloud, without any setup prerequisites or enforced cooperation.

 

Manuscript

 

Bibtex