Gabriel Campero Durand

M.Sc. Gabriel Campero Durand

Institut für Technische und Betriebliche Informationssysteme (ITI)
AG Datenbanken & Software Engineering
Universitätsplatz 2, 39106, Magdeburg, G29-108
Vita

Gabriel Campero Durand is a PhD student and associated researcher at the Databases and Software Engineering group of the Otto-von-Guericke University of Magdeburg.
He received his M.Sc. degree in Data and Knowledge Engineering at the University of Magdeburg in 2017. Before his current role he worked with IBM Research, and IBM Cloud Availability Monitoring in Boblingen.
His research focuses on production-ready applications of AI techniques to data management, with a focus on deep reinforcement learning.

Projekte

Completed projects

COOPeR: Cross-device OLTP/OLAP PRocessing
Duration: 01.09.2016 bis 30.06.2021

Database Management Systems (DBMS) face two challenges today. On the one hand DBMS must handle Online Transaction Processing (OLTP), and Online Analytical Processing (OLAP) in combination to enable real-time analysis of business processes. The real-time analysis of business processes is necessary to improve the quality of reports and analyzes, since fresh data is favored for modern analysis rather than historical data only. On the other hand, computer systems become increasingly heterogeneous to provide better hardware performance. The architecture changes from single-core CPUs to multi-core CPUs supported by several co-processors. These trends must be considered in DBMS to improve the quality and performance, and to ensure that DBMS satisfy future requirements (e.g., more complex queries, or more increased data volume). Unfortunately, current research approaches address only one of these two challenges: either the combination of OLTP and OLAP workloads in traditional CPU-based systems, or co-processor acceleration for a single workload type is considered. Therefore, an unified approach addressing both challenges at once is missing. In this project we want to include both challenges of DBMS to enable efficient processing of combined OLTP / OLAP workloads in hybrid CPU / Co-processor systems. This is necessary in order to realize real-time business intelligence. The main challenge is guaranteeing the ACID properties for OLTP, while at the same time to combine and to process efficiently OLTP / OLAP workloads in such a hybrid systems.

View project in the research portal

Forschung
  • Learning-augmented (& production-ready) solutions for data management
    • Data partitioning
    • Join order optimitzation (cardinality estimation and plan-space search improvements)
    • ...
  • Management of machine learning in production
    • Safety
    • Interpretability
    • Tooling
    • Curricula design and sample efficiency
    • Data management for ML
    • ...
  • Network analysis

Last Modification: 31.03.2021 - Contact Person: