ABAlexander Baumstark

Alexander Baumstark

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Alexander Baumstark has a total of 12 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Let the Hardware do it: Fast Graph Query processing through FPGA-accelerated Late Materialization and Decompression

    by Lei Chen, Alexander Baumstark, Andreas Becher, Daniel Ziener, Kai-Uwe Sattler

    Property access is a dominant performance bottleneck in modern graph DBMSs, especially for analytical queries involving multi-hop traversals and selective projection. We observe that premature materialization and repeated property decoding cause excessive data movement and irregular memory access, limiting throughput and scalability. While Late Materialization (LM) has proven effective in column stores, its adoption in graph engines is hindered by irregular access patterns and fine-grained decoding costs that are poorly handled by general-purpose CPUs, where cache misses, branch divergence, and synchronization overhead limit effective memory bandwidth utilization. We, therefore, propose an FPGA-accelerated co-processing approach that tightly integrates compression-aware LM into the graph query pipeline. By exploiting the deterministic and streaming-friendly nature of FSST decoding, we offload string decompression to a deeply pipelined hardware dataflow engine, enabling high throughput. Our experimental results demonstrate that the hardware-accelerated approach achieves speedups of up to 24.6 × over a CPU-only baseline.

  • Open Access

    Characterizing Multi-Host CXL 2.0: Latency, Bandwidth, and Synchronization Costs

    by Laurin Martins, Alexander Baumstark, Andreas Becher, Kai-Uwe Sattler

    Compute Express Link (CXL) is an interconnect that extends the traditional memory hierarchy beyond a single machine. It enables disaggregated memory architecture, where memory can be pooled and shared across multiple hosts effectively. Previous systems and approaches were limited to CXL 1.0 which is restricted to intra-host scaling of memory resources with no out of the box support for sharing across multiple hosts. We evaluate a commercially available CXL 2.0 memory switching platform for multi-host shared memory. While prior work has largely focused on emulation, FPGA-based or non-available prototypes, we investigate the real-world performance and system implications of deploying CXL as shared memory across multiple hosts on a commercially available setup. Our evaluation reveals fundamental trade-offs: Access to shared CXL memory exhibits up to 4x higher latency and significantly lower bandwidth compared to local DRAM, but enables flexible memory pooling with fair sharing across hosts. More importantly, we show that the lack of hardware cache coherence in CXL 2.0 introduces non-trivial challenges, requiring carefully chosen instruction patterns.


  • Open Access

    GraphRTX: Lighting the Way to Scalable Graph Analytics

    by Alexander Baumstark, Kai-Uwe Sattler

    Modern GPU architectures are powerful platforms for accelerating database processing, including graph analytics. However, due to irregular access patterns in graphs caused by properties such as power-law distribution of vertex degrees, pure SMT execution models can result in resource under-utilization and poor memory coalescing. Modern GPUs address this issue by using heterogeneous accelerators for different types of workload. Among these are ray tracing (RT) cores, which perform hardware-accelerated spatial index traversal, overcoming similar challenges of irregular memory access. By offloading traversal and intersection to fixed-function hardware, RT cores align well with graph traversal patterns. However, there is still no unified graph representation for RT that avoids costly index rebuilding across algorithms. To address this gap, we introduce GraphRTX, a framework that leverages the RT cores of modern NVIDIA GPUs to accelerate graph algorithms. GraphRTX employs a unified graph representation that maps graphs to bounding volume hierarchies (BVHs) and a ray-based query abstraction and enables multiple algorithms to operate on the same model. We also propose a hybrid execution model that combines RT acceleration with GPU thread-level parallelism. Experiments on real-world datasets demonstrate that GraphRTX matches or outperforms state-of-the-art baselines, offering lower BVH (re)build overhead and a smaller memory footprint.

  • Open Access

    Uncore your Queries: Towards CPU-less Query Processing

    by Alexander Baumstark, Laurin Martins, Kai-Uwe Sattler


  • Open Access

    Hybrid Transactional/Analytical Graph Processing in Modern Memory Hierarchies

    by Alexander Baumstark, Muhammad Attahir Jibril, Kai-Uwe Sattler

    Today’s enterprise computing architectures are characterized by a complex memory hierarchy: different application requirements in terms of latency, bandwidth, persistence, and access pattern, as well as characteristics of available memory and storage technology require combining different technologies. Building highly efficient data management and analytics solutions that meet the challenges of modern applications requires to utilize this memory hierarchy, i.e., taking the specific characteristics of a given technology into account and keeping data objects in the optimal level. In this paper, we present results of our project on exploiting modern memory hierarchies in support of hybrid transactional/analytical processing (HTAP) on graph data. First, we discuss the design and evaluation of data structures and query operations for graph data in persistent memory. Second, we present an approach to support the analysis of graph data on GPU-based accelerators with dedicated memory by efficient data transfer and consistency mechanisms. Finally, we propose a storage and processing strategy for (bi-)temporal graphs using temporal materialized views while exploiting the memory hierarchy.


  • Open Access

    Temporal graph processing in modern memory hierarchies

    by Alexander Baumstark, Muhammad Attahir Jibril, Kai-Uwe Sattler


  • Open Access

    So Far and yet so Near - Accelerating Distributed Joins with CXL

    by Alexander Baumstark, Marcus Paradies, Kai-Uwe Sattler, Steffen Kläbe, Stephan Baumann

    Distributed partitioned joins are one of the most expensive operators in distributed DBMSs where a major part of the execution is attributed to network transfer costs. Although high-speed network technologies, such as RDMA, can lower this cost, they still come with significantly higher latency than local DRAM access. The emerging CXL interconnect protocol promises to provide direct and cache-coherent access to remote memory while offering byte-addressable memory access without CPU intervention. For short-distance communication in distributed DBMSs, CXL represents an interesting alternative for low-latency requirements. In this work, we explore how CXL can be leveraged for engine-internal communication and data exchange. We discuss and apply communication strategies to distributed joins. We emulate various CXL characteristics based on optimistic and pessimistic assumptions on the real performance of upcoming CXL devices and evaluate their impact on the execution of distributed joins. Our results show that CXL has the potential to improve distributed join performance.


  • Open Access

    Accelerating Large Table Scan Using Processing-In-Memory Technology

    by Alexander Baumstark, Muhammad Attahir Jibril, Kai-Uwe Sattler

    Today’s systems are capable of storing large amounts of data in main memory. Particularly, in-memory DBMSs benefit from this development. However, the processing of data from the main memory necessarily has to run via the CPU. This creates a bottleneck, which affects the possible performance of the DBMS. Processing-In-Memory (PIM) is a paradigm to overcome this problem, which was not available in commercial systems for a long time. With the availability of UPMEM, a commercial product is finally available that provides PIM technology in hardware. In this work, we focus on the acceleration of the table scan, a fundamental database query operation. We show and investigate an approach that can be used to optimize this operation by using PIM. We evaluate the PIM scan in terms of parallelism and execution time in benchmarks with different table sizes and compare it to a traditional CPU-based table scan. The result is a PIM table scan that outperforms the CPU-based scan significantly.


  • Temporal Graph Processing in Modern Memory Hierarchies

    by Alexander Baumstark, Muhammad Attahir Jibril, Kai-Uwe Sattler


  • Open Access

    Processing-in-Memory for Databases: Query Processing and Data Transfer

    by Alexander Baumstark, Muhammad Attahir Jibril, Kai-Uwe Sattler

    The Processing-in-Memory (PIM) paradigm promises to accelerate data processing by pushing down computation to memory, reducing the amount of data transfer between memory and CPU, and – in this way – relieving the CPU from processing. Particularly, in in-memory databases memory access becomes a performance bottleneck. Thus, PIM seems to offer an interesting solution for database processing. In this work, we investigate how commercially available PIM technology can be leveraged to accelerate query processing by offloading (parts of) query operators to memory. Furthermore, we show how to address the problem of limited PIM storage capacity by interleaving transfer and computation and present a cost model for the data placement problem.