In-Memory DBMS for High-Performance Memories
Eduardo Cunha de Almeida, UFPR - Departamento de Informática
In this talk, we discuss query processing in Processor-in-Memory (PIM) and Non-Uniform Memory Access (NUMA) memory architectures in order to tackle the memory wall problem. This problem has a direct impact on the execution performance and power consumption of database systems due to the movement of large volumes of data through the memory hierarchy. This data movement is the typical case of complex data processing, such as Online Analytical Processing (OLAP) and data cleansing processes, which need to execute in the CPU operations such as filters, projections and joins. In this talk, we discuss the work that our group has been doing to mitigate the effect of such data movement. We discuss some of our contributions and future work in the coordination of query processing in PIM and NUMA architectures, and the necessary hardware modifications to enable such processing. Our talk highlights the results that we recently published in database and computer architecture conferences, such as VLDB, ICDE, DATE, CIKM, ADMS@VLDB and Damon@SIGMOD.

Please email for a Zoom link
About the speaker
Eduardo Cunha de Almeida received the Ph.D. degree in computer science from the University of Nantes (AtlasGDD INRIA Team), France, in 2009. He joined the Federal University of Paraná (UFPR), Brazil, in 2009, where he is currently an Associate Professor at the Department of Informatics. His research interests include database systems (query processing and optimization) and data management (data quality, data profiling and data cleansing). He was a Research Associate at the Interdisciplinary Centre for Security, Reliability and Trust (SnT), Luxembourg, from 2013 to 2015. He was an Engineer of Database Technology at HSBC Bank (now Bradesco) and GVT Telecom (now Vivo), Brazil, from 1998 to 2005.
Date & Time
Thursday, August 11, 2022 - 14:00