Title: Data Mining Meets Systems - tools and case studies

Speaker: Christos Faloutsos, CMU/SCS

Abstract
How can we model bursty disk or network traffic?
How can we monitor a large data center like self-*,
and spot anomalies?
We show how to answer these questions using some powerful
tools from data mining, like self-similarity, power laws,
singular value decomposition and tensors.
We also give case studies with real datasets,
like the HP disk trace data and self-* measurement data
(`InteMon').

The overarching message is that the two fields have a *lot*
to offer to each other: Data Mining can provide tools for
monitoring, design and anomaly detection of large data centers;
systems offer powerful tools to help analyze Tera- and Peta-byte
size datasets, that are impossible to study otherwise.