Modern cloud computing data centers typically deliver virtualized services to tens of thousands of cloud users continuously, using hundreds or even thousands of commodity servers. Monitoring the availability, performance, and security of this data center infrastructure efficiently and effectively in real time is a daunting task, requiring significant processing and storage resources.
Xu et al. present their lightweight, power-series-based streaming frequency traffic metric that uses aggregated Internet protocol (IP) traffic streams to uncover real-time anomalous traffic patterns on Internet data center servers without the large processing and storage overheads of traditional monitoring methods. The theoretical basis for the streaming frequency metric is discussed and compared with traditional methods that have significantly higher resource and processing requirements.
An analysis and discussion of the properties of the metric are presented, as well as a demonstration using example data collected from large data centers, to show the application of the metric in detecting anomalous traffic, aiding network management and security monitoring of data center servers.
Applications for real-time data center network monitoring are considered, conclusions and future work are discussed, and thorough references are provided. An interesting potential tool for efficient, high-level, real-time monitoring of large cloud infrastructures that could be used to highlight components deserving closer scrutiny with more detailed security monitoring tools.