ATOM: Automated tracking, orchestration and monitoring of resource usage in infrastructure as a service systems

We present ATOM, an efficient and effective framework to enable automated tracking, monitoring, and orchestration of resource usage in an Infrastructure as a Service (IaaS) system. We design a novel tracking method to continuously track important performance metrics with low overhead, and develop a principal component analysis (PCA) based approach with quality guarantees to continuously monitor and automatically find anomalies based on the approximate tracking results. Lastly, when potential anomalies are identified, we use introspection tools to perform memory forensics on virtual machines (VMs) to identify malicious behavior inside a VM. We deploy ATOM in an IaaS system to monitor VM resource usage, and to detect anomalies. Various attacks are used as examples to demonstrate how ATOM is both effective and efficient to track and monitor resource usage, detect anomalies, and orchestrate system resource usage.

我们提出了ATOM，这是一个高效且有效的框架，用于在基础设施即服务（IaaS）系统中实现资源使用情况的自动跟踪、监测与编排。我们设计了一种新颖的跟踪方法，能够以较低开销持续跟踪重要性能指标，并开发了一种基于主成分分析（PCA）且有质量保证的方法，以便依据近似跟踪结果持续监测并自动发现异常情况。最后，当识别出潜在异常时，我们会使用内省工具对虚拟机（VM）执行内存取证，以识别虚拟机内部的恶意行为。我们将ATOM部署在IaaS系统中，用于监测虚拟机资源使用情况并检测异常。我们以各种攻击为例，展示ATOM如何高效且有效地跟踪和监测资源使用情况、检测异常以及编排系统资源使用。