Automated Dependency Discovery of Hosts and Network Services in Distributed Systems
In a modern enterprise network of scale, dependencies between hosts and network services are surprisingly complex, typically undocumented, and rarely static.
The automated discovery of these dependencies is a current challenge for network management and troubleshooting.
In this talk I will present a method for addressing this challenge. The method is based on learning the parameters of a generative probabilistic model using little more than timings of packet transmission and reception, and then performing statistical hypothesis testing on the components of the model.
I will also show results from applying this approach to real data from a trace collecting network events at Microsoft Research Cambridge.
Joint work with Alex Simma, John MacCormick, Paul Barham, Richard Black, Rebecca Isaacs and Richard Mortier- ACM Bay Area Chapter
Moises Goldszmidt is a principal researcher with Microsoft Research in the Silicon Valley Campus. Prior to Microsoft, Moises held similar positions with Hewlett-Packard Labs, SRI International, and Rockwell Science Center, and was a principal scientist with Peakstone Corporation (start-up).
His research interests include probabilistic reasoning, graphical models, pattern recognition, statistical induction, machine learning, and artificial intelligence. Dr. Goldszmidt has a PhD degree in computer science from UCLA (1992).
Moises Goldszmidt conducts a machine learning investigation which answers the question "Why Can't I Log in to my E-mail?"
Goldszmidt answers this questions by explaining that it is most likely not the server that is a problem but one of many in a "host of services" such as "authentication and anti-virus systems and active directories."