El próximo 22 de julio de 2016 a las 11:30h en la Sala de Grados de la Facultad de Informática de la Universidad Complutense de Madrid, Frans Slothouber impartirá la siguiente conferencia: Control Center CPU capacity planning with Big Data based predictive analytics.
Descarga la presentación de la ponencia
Puedes visualizar y descargar la presentación de la ponencia en el siguiente enlace.
Train dispatchers in the Netherlands use a large heterogeneous stack of software applications to control train traffic. This stack runs on a cluster of computers and uses a limited resource, CPU time. Proper provisioning of this resource is needed because failure to do so can lead to a sluggish system that disrupts operations and consequently the flow of trains. In this talk we describe how a statistical model can be build and applied to support this kind capacity planning. The model was constructed using two big data sets with historical data collected from the 13 control centers that control the 13 areas the Dutch railway network is partitioned into. The first set contains data from a performance monitoring tool that collects data on minute by minute CPU time usage. The second set contains all logfiles of the applications in the stack.
Frans Slothouber works as Data Scientist after working as System Architect, Consultant and Senior Technical Designer among other positions in the Dutch Railway Industry. His education is M. Sc. Computer Science graduated in 1994 at Twente University. His expertise is performance analysis, predictive analytics, data visualization, network design, machine learning and software development among others. From 1995, he has been involved in the following projects: OMA Data Analysis, AKP Predictive Analysis, Machine Learning Based Logfile Analysis, MTIS Automated Testing, OTIS Automated Testing, Trend Analysis, Wide Area Signaling Network, Railway Simulator, Signalling Converter, Railway Decision Support System and Network Performance Analysis. He also has written several research publications in conference proceedings.