日時：2016年7月1日(金) 15:00 – 16:30
題目: Monitorology and Big Data in the Age of Data Analytics
講師: Prof. Miroslaw Malek (Universita della Svizzera italiana)
We focus on the art of observing the world by humans and electronic devices such as sensors and meters that, in general, we call monitors. We then move to monitoring devices, define main monitoring objectives and pose five challenges for effective and efficient monitoring that still need a lot of research. In the age of computricity, where compute power like electricity is easily available and easy to use across the globe, and big data that is generated in enormous amounts and ever-increasing rates, the question, what to monitor and how, will become ever more relevant to save the world from flood of meaningless, dumb data, leading frequently to false conclusions and wrong decisions whose impact may range from a minor inconvenience to loss of lives and major disasters.
We argue, that in the age of Big Data, current complexity levels and necessity of dealing with time, in addition to classical synthesis and analysis methods, we need to turn to empirical data-driven approaches using data analytics to, for example, proactive fault management which require monitoring, online measurement, online analysis, diagnosis, failure prediction and decision making to support recovery and nonstop computing and communication. To illustrate such approaches two case studies are presented: In the first case study, we address the problem of proactive fault management by demonstrating how runtime monitoring, variable selection and model re-evaluation lead to effective failure prediction.
We also present the quality analysis of such prediction to determine whether it results in dependability gain. The second case study illustrates how by observation and measurement of CPU and memory features a malicious software (malware) can be detected on line. Finally, we conclude that models derived from monitoring and measurement will continue gaining on significance and impact and list the major challenges for data-driven research on dependability and security.
広島市立大学 角田良明 TEL:082-830-1696
E-mail: kakuda&hiroshima-cu.ac.jp (送信時に&を@に換えてください）