Embedded Files
Target of the course:
Over the last decade, we have witnessed the emergence of data-intensive applications that need to handle very large flows of data. Examples of such applications include network monitoring, financial applications, security log, and sensors applications. For all these applications, there is a growing need for (new) techniques capable of monitoring or analyzing these streams to detect outliers, intrusions, unusual or anomalous activities, complex correlations, extreme events, or the emergence of patterns. These techniques are grouped together under the term “ Big data stream analytics”. These techniques must be capable of processing the input data quick enough to keep pace with the rate of the stream. The solution adopted to process such data streams is to trade off accuracy for space. Actually, one of the specificities of data intensive applications is that they do not require accurate responses to queries, only (high quality) approximate responses computed by summarizing the data are acceptable. Streaming algorithms, through synopses, handle such features. They read their input data sequentially, and there is no requirement regarding the order in which data items are received. They use working memory whose size is much smaller than both the input size, and the domain from which items are drawn. There is a large panel of streaming algorithms that vary according to the number of passes they need to process their input process, the size of the memory they use, the time needed to process each read item, or whether they are randomized or deterministic.
This teaching plan aims at proposing a comprehensive survey of these techniques, across several practical usage (Distributed system safety, network security, cloud monitoring, ...).
Keywords:
Similarity metrics, Distributed monitoring: the need of sampling. Enhancement of a core sketching algorithm, Usage of stream analysis methods for Stream processing. Performance analysis of studied algorithms (Lab activity).
Course material:
Google Sites
Report abuse