In Partial Fulfillment of the Requirements for the Degree of Master of Science
will defend her thesis
Disk Failure Prediction in Heterogeneous Environments Using Deep Neural Networks
Hard drive failures at data centers may cause temporary system unavailability or even complete data loss. Thus, foreseeing the failure and migrating data as well as services out of dangerous disks beforehand are of critical importance. The goal of this research is to build a predictive model for disk failure in heterogeneous environments with high accuracy, a high false detection rate and a reasonable false alarm rate. We propose a predictive framework consisting of a baseline predictor for all models of different manufactures, and a core predictor which is designed separately for different models of different manufactures. The baseline predictor is used in case there’s a lack of historical data for the new model of a specific disk manufacturer. These two predictors made use of 6 original SMART features, and 18 newly-defined features based on these 6 original SMART features.
Date: Wednesday, April 24, 2019
Time: 1:00 - 2:30 PM
Place: PGH 501D
Advisors: Dr. Jehan-François Pâris
Faculty, students, and the general public are invited.