Abstract

We look into completely modelling a storage-based data centre and find that traditional performance models are too brittle. Simply put, a brittle model is often inaccurate and incorrect. That turns off systems people and administrators who then opt for simple, but often inaccurate, “rules-of-thumb”. After learning lessons the hard way, we went and implemented a robust modelling framework we call IRON Model. IRON Model leverages the redundancy of high-level system specifications described through (queueing) models and low-level system implementation to localize many types of system-model inconsistencies. IRON Model can guide designers to the potential source of the discrepancy, and, if appropriate, can semi-automatically evolve the models (using statistical techniques) to handle unanticipated inputs. This is largely a systems talk and touches on the minimal mathematical tools needed to build robust models within it, which then help manage the system better. This is joint work with CMU .

Bio: I am a systems person, currently with a focus on file systems and storage technologies and high-performance data centers. I also have great interest in applying machine learning and queuing analysis to help simplify and automate system management. Since September 2007 I have been a Researcher at Microsoft Research in Cambridge, UK. I received my PhD/MS/BS from Carnegie Mellon.