Theodoros Gkountouvas

"Improving Data Analysis by Exploiting Temporal Information"

( MPI-SWS talk in Kooperation mit dem Fachbereich Informatik)

Internet of Things (IoT) applications produce large amount of sensor data. This data contains temporal information that can improve data analysis if utilized correctly. This talk is divided in three separate parts. Initially, we will discuss about examples of IoT applications and an overview of IoT platforms. Then, I will go deeper and show Freeze-Frame File System (FFFS), which allows temporal queries on demand. FFFS provides strong guarantees for users and efficiently handles temporal queries. Finally, we will talk about temporal caching. We effectively exploit temporal information in order to devise more sophisticated eviction policies for data analysis. Better utilization of cache leads to less I/O requests and less computations which improve the deployment of IoT platforms in terms of performance and costs.

Bio: Theodoros Gkountouvas is a Ph.D. candidate at Cornell and his work is focused on Distributed Systems. He has received M.Eng. and M.Sc. Computer Science degrees from Cornell on May 2013 and December 2016 respectively. Theodoros finished his undergraduate studies in Electrical and Computing Engineering from Polytechnic National Technical University of Athens (NTUA). He has worked as an Associate Researcher between September 2011 and May 2012 at CSLab in NTUA. He is currently interested in improving systems for IoT platforms and more broadly in topics that lie in the intersection of Distributed Systems and Machine Learning.