Domains such as financial trading, advertising and marketing have been fertile breeding grounds for cutting edge data-driven applications. Yet the systems that power these services are still running on decades old technology. Traders can run algorithms that make millions of dollars in milliseconds but the underlying systems are still serviced by tools that provide mere searching and graphing. Even the most powerful IT tools only look at small fractions of data, leaving most IT professions to hunt around in the dark. In this talk we will explore the application of machine learning in the domain of IT operations. In the datacenter there are thousands of sources of event data. By modeling the datacenter as a source of multiple data streams, we can apply techniques from other domains, modified to address the environments that now power much of  our economy. Our focus here is around correlated anomaly detection across multiple time series data sets. By filtering streams of events that deviate from previously established patterns we can surface likely correlations of these streams for further exploration and analysis. Anomalies may be surfaced by anything from simple heuristics such as a reduced count of events, reference points such as specific types of event counts crossing Bollinger bands,  or by more complex techniques such as probabilistic graphical models. Attendees will learn how techniques used in other domains of machine learning can be applied to help IT operators keep quality of service high. Attendees should be familiar with either machine learning techniques or with IT operations and data center architecture.