Monalytics

Abstract

To effectively manage large-scale data centers and utility clouds, operators must understand current system and application behaviors. This requires continuous monitoring along with online analysis of the data captured by the monitoring system. As a result, there is a need to move to systems in which both tasks can be performed in an integrated fashion, thereby better able to drive online system management. Coining the term 'monalytics' to refer to the combined monitoring and analysis systems used for managing large-scale data center systems, this paper articulates principles for monalytics systems, describes software approaches for implementing them, and provides experimental evaluations justifying principles and implementation approach. Specific technical contributions include consideration of scalability across both 'space' and 'time', the ability to dynamically deploy and adjust monalytics functionality at multiple levels of abstraction in target systems, and the capability to operate across the range of application to hypervisor layers present in large-scale data center or cloud computing systems. Our monalytics implementation targets virtualized systems and cloud infrastructures, via the integration of its functionality into the Xen hypervisor.