DFG project part of the priority programme 2377: https://spp2377.uos.de/

Abstract:

The overall objective of this project is to enable main-memory database systems to benefit from an elastic memory hierarchy. This approach is groundbreaking not only because it incorporates memory technologies such as NVRAM, HBM, PIM and NMC components, but above all because of a selection process for suitable components and technologies to be used in the elastic memory hierarchy, such as memory compression, GPU acceleration or the use of remote memory resources. The actual configuration of the memory hierarchy for a specific use case will be based on a multi-criteria analysis of the trade-offs between various non-functional parameters of a system. We assume that the system follows the principles of autonomic computing: it adapts to non-functional properties and to changes in the availability of the components of the memory hierarchy.

Principal Investigator

• Prof. Dr. Maximilian E. Schüle

Project Members

• Maha Alwahibi, M.Sc.
• Bachana Jangebashvili, B.Sc.
• Duong Tung Nguyen, M.Sc.
• Lakshmi Nair, B.Sc.

Publications

• Maha Alwahibi, Maximilian E. Schüle: Compression-Aware GPU Buffer Management (Datenbank-Spektrum, 2026)
• Maha Alwahibi, Bachana Jangebashvili, Maximilian E. Schüle: Into the CUDA Multiverse of Runtime Compilation: Exploring Kernel Fusion Strategies for GPU Database Systems (52nd International Conference on Very Large Databases (VLDB 2026))