Peter Boncz
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Peter Boncz has a total of 14 co-authors. Here are the top co-authors by number of shared publications:
- Azim Afroozeh
CWI, Amsterdam, Netherlands
5 shared publications - Leonardo Kuffo
CWI, Amsterdam, Netherlands
4 shared publications - Lotte Felius
CWI, Amsterdam, Netherlands
3 shared publications - Ilaria Battiston
Centrum Wiskunde & Informatica, Amsterdam, The Netherlands
2 shared publications - Bogdan Nicuşor Bindea
Department of Computer Science, Technical University of Cluj-Napoca, 400027 Cluj-Napoca, Romania
1 shared publication - Elena Krippner
CWI, Amsterdam, Netherlands
1 shared publication
Latest Publications
- Open Access
- 22 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-8
Bang for the Buck: Vector Search on Cloud CPUs
- 22 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-8
Open AccessBang for the Buck: Vector Search on Cloud CPUs
- Open Access
- 22 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-10
G-ALP: Rethinking Light-weight Encodings for GPUs
- 22 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-10
Open AccessG-ALP: Rethinking Light-weight Encodings for GPUs
- Open Access
- 17 June 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Management of Data
- Vol. 3 (3) , 1-26
PDX: A Data Layout for Vector Similarity Search
We propose Partition Dimensions Across (PDX), a data layout for vectors (e.g., embeddings) that, similar to PAX [6], stores multiple vectors in one block, using a vertical layout for the dimensions (Figure 1). PDX accelerates exact and approximate similarity search thanks to its dimension-by-dimension search strategy that operates on multiple-vectors-at-a-time in tight loops. It beats SIMD-optimized distance kernels on standard horizontal vector storage (avg 40% faster), only relying on scalar code that gets auto-vectorized. We combined the PDX layout with recent dimension-pruning algorithms ADSampling [19] and BSA [52] that accelerate approximate vector search. We found that these algorithms on the horizontal vector layout can lose to SIMD-optimized linear scans, even if they are SIMD-optimized. However, when used on PDX, their benefit is restored to 2-7x. We find that search on PDX is especially fast if a limited number of dimensions has to be scanned fully, which is what the dimension-pruning approaches do. We finally introduce PDX-BOND, an even more flexible dimension-pruning strategy, with good performance on exact search and reasonable performance on approximate search. Unlike previous pruning algorithms, it can work on vector data ''as-is'' without preprocessing; making it attractive for vector databases with frequent updates.
- 17 June 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Management of Data
- Vol. 3 (3) , 1-26
Open AccessPDX: A Data Layout for Vector Similarity Search
We propose Partition Dimensions Across (PDX), a data layout for vectors (e.g., embeddings) that, similar to PAX [6], stores multiple vectors in one block, using a vertical layout for the dimensions (Figure 1). PDX accelerates exact and approximate similarity search thanks to its dimension-by-dimension search strategy that operates on multiple-vectors-at-a-time in tight loops. It beats SIMD-optimized distance kernels on standard horizontal vector storage (avg 40% faster), only relying on scalar code that gets auto-vectorized. We combined the PDX layout with recent dimension-pruning algorithms ADSampling [19] and BSA [52] that accelerate approximate vector search. We found that these algorithms on the horizontal vector layout can lose to SIMD-optimized linear scans, even if they are SIMD-optimized. However, when used on PDX, their benefit is restored to 2-7x. We find that search on PDX is especially fast if a limited number of dimensions has to be scanned fully, which is what the dimension-pruning approaches do. We finally introduce PDX-BOND, an even more flexible dimension-pruning strategy, with good performance on exact search and reasonable performance on approximate search. Unlike previous pruning algorithms, it can work on vector data ''as-is'' without preprocessing; making it attractive for vector databases with frequent updates.
- Open Access
- 21 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 51-54
Reproducibility Report for ACM SIGMOD 2024 Paper: 'ALP: Adaptive Lossless Floating-Point Compression'
- 21 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 51-54
Open AccessReproducibility Report for ACM SIGMOD 2024 Paper: 'ALP: Adaptive Lossless Floating-Point Compression'
- Open Access
- 9 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-11
Accelerating GPU Data Processing using FastLanes Compression
We show that compression can be a win-win for GPU data processing: it not only allows to store more data in GPU global memory, but can also accelerate data processing. We show that the complete redesign of compressed columnar storage in FastLanes, with its fully data-parallel bit-packing and encodings, also benefits GPU hardware. We micro-benchmark the performance of FastLanes on two GPU architectures (Nvidia T4 and V100) and integrate FastLanes in the Crystal GPU query processing prototype. Our experiments show that FastLanes decompression significantly outperforms previous decompression methods in micro-benchmarks, and can make end-to-end SSB queries up to twice faster compared to uncompressed query processing - in contrast to previous work where GPU decompression caused execution to slow down. We further discovered that an access granularity of decoding vectors of 1024 values is too large for a single GPU warp due to register pressure. We mitigate this here using mini-vectors - a future work question is how to further reduce this granularity with minimal impact on efficiency.
- 9 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-11
Open AccessAccelerating GPU Data Processing using FastLanes Compression
We show that compression can be a win-win for GPU data processing: it not only allows to store more data in GPU global memory, but can also accelerate data processing. We show that the complete redesign of compressed columnar storage in FastLanes, with its fully data-parallel bit-packing and encodings, also benefits GPU hardware. We micro-benchmark the performance of FastLanes on two GPU architectures (Nvidia T4 and V100) and integrate FastLanes in the Crystal GPU query processing prototype. Our experiments show that FastLanes decompression significantly outperforms previous decompression methods in micro-benchmarks, and can make end-to-end SSB queries up to twice faster compared to uncompressed query processing - in contrast to previous work where GPU decompression caused execution to slow down. We further discovered that an access granularity of decoding vectors of 1024 values is too large for a single GPU warp due to register pressure. We mitigate this here using mini-vectors - a future work question is how to further reduce this granularity with minimal impact on efficiency.
- Open Access
- 9 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-5
DuckDB-SGX2: The Good, The Bad and The Ugly within Confidential Analytical Query Processing
We provide an evaluation of an analytical workload in a confidential computing environment, combining DuckDB with two technologies: modular columnar encryption in Parquet files (data at rest) and the newest version of the Intel SGX Trusted Execution Environment (TEE), providing a hardware enclave where data in flight can be (more) securely decrypted and processed. One finding is that the "performance tax" for such confidential analytical processing is acceptable compared to not using these technologies. We eventually manage to run TPC-H SF30 with under 2x overhead compared to non-encrypted, non-enclave execution; we show that, specifically, columnar compression and encryption are a good combination. Our second finding consists of dos and don'ts to tune DuckDB to work effectively in this environment. There are various performance hazards: potentially 5x higher cache miss costs due to memory encryption inside the enclave, NUMA penalties, and highly elevated cost of swapping pages in and out of the enclave - which is also triggered indirectly by using a non-SGX-aware malloc library.
- 9 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-5
Open AccessDuckDB-SGX2: The Good, The Bad and The Ugly within Confidential Analytical Query Processing
We provide an evaluation of an analytical workload in a confidential computing environment, combining DuckDB with two technologies: modular columnar encryption in Parquet files (data at rest) and the newest version of the Intel SGX Trusted Execution Environment (TEE), providing a hardware enclave where data in flight can be (more) securely decrypted and processed. One finding is that the "performance tax" for such confidential analytical processing is acceptable compared to not using these technologies. We eventually manage to run TPC-H SF30 with under 2x overhead compared to non-encrypted, non-enclave execution; we show that, specifically, columnar compression and encryption are a good combination. Our second finding consists of dos and don'ts to tune DuckDB to work effectively in this environment. There are various performance hazards: potentially 5x higher cache miss costs due to memory encryption inside the enclave, NUMA penalties, and highly elevated cost of swapping pages in and out of the enclave - which is also triggered indirectly by using a non-SGX-aware malloc library.
- Open Access
- 9 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 516-519
OpenIVM: a SQL-to-SQL Compiler for Incremental Computations
This demonstration presents a new Open Source SQL-to-SQL compiler for Incremental View Maintenance (IVM). While previous systems, such as DBToaster, implemented computational functionality for IVM in a separate system, the core principle of OpenIVM is to make use of existing SQL query processing engines and perform all IVM computations via SQL. This approach enables the integration of IVM in these systems without code duplication. Also, it eases its use in cross-system IVM, i.e. to orchestrate an HTAP system in which one (OLTP) DBMS provides insertions/updates/deletes (deltas), which are propagated using SQL into another (OLAP) DBMS, hosting materialized views. Our system compiles view definitions into SQL to eventually propagate deltas into the table that materializes the view, following the principles of DBSP. Under the hood, OpenIVM uses the DuckDB library to compile (parse, transform, optimize) the materialized view maintenance logic. We demonstrate OpenIVM in action (i) as the core of a DuckDB extension module that adds IVM functionality to it and (ii) powering cross-system IVM for HTAP, with PostgreSQL handling updates on base tables and DuckDB hosting materialized views on these.
- 9 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 516-519
Open AccessOpenIVM: a SQL-to-SQL Compiler for Incremental Computations
This demonstration presents a new Open Source SQL-to-SQL compiler for Incremental View Maintenance (IVM). While previous systems, such as DBToaster, implemented computational functionality for IVM in a separate system, the core principle of OpenIVM is to make use of existing SQL query processing engines and perform all IVM computations via SQL. This approach enables the integration of IVM in these systems without code duplication. Also, it eases its use in cross-system IVM, i.e. to orchestrate an HTAP system in which one (OLTP) DBMS provides insertions/updates/deletes (deltas), which are propagated using SQL into another (OLAP) DBMS, hosting materialized views. Our system compiles view definitions into SQL to eventually propagate deltas into the table that materializes the view, following the principles of DBSP. Under the hood, OpenIVM uses the DuckDB library to compile (parse, transform, optimize) the materialized view maintenance logic. We demonstrate OpenIVM in action (i) as the core of a DuckDB extension module that adds IVM functionality to it and (ii) powering cross-system IVM for HTAP, with PostgreSQL handling updates on base tables and DuckDB hosting materialized views on these.
- Open Access
- 9 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 4
Making Data Management Better with Vectorized Query Processing
Vectorized query processing is a query processing technique that 20 years ago was introduced in the Vectorwise database system, originally developed in the CWI database architectures research group. Nowadays, it is employed by most analytical data systems, from recent systems like DuckDB, Clickhouse, Velox, DataFusion and Polars to analytical cloud services such as Snowflake, BigQuery, Databricks and MotherDuck. The keynote will tell the story of vectorized query processing, as it evolved in research and industry over the past two decades. It will conclude with a reflection on the purpose of our research ("making data management better").
- 9 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 4
Open AccessMaking Data Management Better with Vectorized Query Processing
Vectorized query processing is a query processing technique that 20 years ago was introduced in the Vectorwise database system, originally developed in the CWI database architectures research group. Nowadays, it is employed by most analytical data systems, from recent systems like DuckDB, Clickhouse, Velox, DataFusion and Polars to analytical cloud services such as Snowflake, BigQuery, Databricks and MotherDuck. The keynote will tell the story of vectorized query processing, as it evolved in research and industry over the past two decades. It will conclude with a reflection on the purpose of our research ("making data management better").
- Top Cited PapersOpen Access
- 8 December 2023
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Management of Data
- Vol. 1 (4) , 1-26
ALP: Adaptive Lossless floating-Point Compression
IEEE 754 doubles do not exactly represent most real values, introducing rounding errors in computations and [de]serialization to text. These rounding errors inhibit the use of existing lightweight compression schemes such as Delta and Frame Of Reference (FOR), but recently new schemes were proposed: Gorilla, Chimp128, PseudoDecimals (PDE), Elf and Patas. However, their compression ratios are not better than those of general-purpose compressors such as Zstd; while [de]compression is much slower than Delta and FOR. We propose and evaluate ALP, that significantly improves these previous schemes in both speed and compression ratio (Figure 1). We created ALP after carefully studying the datasets used to evaluate the previous schemes. To obtain speed, ALP is designed to fit vectorized execution. This turned out to be key for also improving the compression ratio, as we found in-vector commonalities to create compression opportunities. ALP is an adaptive scheme that uses a strongly enhanced version of PseudoDecimals [31] to losslessly encode doubles as integers if they originated as decimals, and otherwise uses vectorized compression of the doubles' front bits. Its high speeds stem from our implementation in scalar code that auto-vectorizes, using building blocks provided by our FastLanes library [6], and an efficient two-stage compression algorithm that first samples row-groups and then vectors.
- 8 December 2023
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Management of Data
- Vol. 1 (4) , 1-26
Top Cited PapersOpen AccessALP: Adaptive Lossless floating-Point Compression
IEEE 754 doubles do not exactly represent most real values, introducing rounding errors in computations and [de]serialization to text. These rounding errors inhibit the use of existing lightweight compression schemes such as Delta and Frame Of Reference (FOR), but recently new schemes were proposed: Gorilla, Chimp128, PseudoDecimals (PDE), Elf and Patas. However, their compression ratios are not better than those of general-purpose compressors such as Zstd; while [de]compression is much slower than Delta and FOR. We propose and evaluate ALP, that significantly improves these previous schemes in both speed and compression ratio (Figure 1). We created ALP after carefully studying the datasets used to evaluate the previous schemes. To obtain speed, ALP is designed to fit vectorized execution. This turned out to be key for also improving the compression ratio, as we found in-vector commonalities to create compression opportunities. ALP is an adaptive scheme that uses a strongly enhanced version of PseudoDecimals [31] to losslessly encode doubles as integers if they originated as decimals, and otherwise uses vectorized compression of the doubles' front bits. Its high speeds stem from our implementation in scalar code that auto-vectorizes, using building blocks provided by our FastLanes library [6], and an efficient two-stage compression algorithm that first samples row-groups and then vectors.
- Top Cited PapersOpen Access
- 1 May 2023
- Published by Association for Computing Machinery (ACM) in Proceedings of the VLDB Endowment
- Vol. 16 (9) , 2132-2144
The FastLanes Compression Layout: Decoding > 100 Billion Integers per Second with Scalar Code
The open-source FastLanes project aims to improve big data formats, such as Parquet, ORC and columnar database formats, in multiple ways. In this paper, we significantly accelerate decoding of all common Light-Weight Compression (LWC) schemes: DICT, FOR, DELTA and RLE through better data-parallelism. We do so by re-designing the compression layout using two main ideas: (i) generalizing the value interleaving technique in the basic operation of bit-(un)packing by targeting a virtual 1024-bits SIMD register, (ii) reordering the tuples in all columns of a table in the same Unified Transposed Layout that puts tuple chunks in a common "04261537" order (explained in the paper); allowing for maximum independent work for all possible basic SIMD lane widths: 8, 16, 32, and 64 bits. We address the software development, maintenance and future-proofness challenges of increasing hardware diversity, by defining a virtual 1024-bits instruction set that consists of simple operators supported by all SIMD dialects; and also, importantly, by scalar code. The interleaved and tuple-reordered layout actually makes scalar decoding faster, extracting more data-parallelism from today's wide-issue CPUs. Importantly, the scalar version can be fully auto-vectorized by modern compilers, eliminating technical debt in software caused by platform-specific SIMD intrinsics. Micro-benchmarks on Intel, AMD, Apple and AWS CPUs show that FastLanes accelerates decoding by factors (decoding >40 values per CPU cycle). FastLanes can make queries faster, as compressing the data reduces bandwidth needs, while decoding is almost free.
- 1 May 2023
- Published by Association for Computing Machinery (ACM) in Proceedings of the VLDB Endowment
- Vol. 16 (9) , 2132-2144
Top Cited PapersOpen AccessThe FastLanes Compression Layout: Decoding > 100 Billion Integers per Second with Scalar Code
The open-source FastLanes project aims to improve big data formats, such as Parquet, ORC and columnar database formats, in multiple ways. In this paper, we significantly accelerate decoding of all common Light-Weight Compression (LWC) schemes: DICT, FOR, DELTA and RLE through better data-parallelism. We do so by re-designing the compression layout using two main ideas: (i) generalizing the value interleaving technique in the basic operation of bit-(un)packing by targeting a virtual 1024-bits SIMD register, (ii) reordering the tuples in all columns of a table in the same Unified Transposed Layout that puts tuple chunks in a common "04261537" order (explained in the paper); allowing for maximum independent work for all possible basic SIMD lane widths: 8, 16, 32, and 64 bits. We address the software development, maintenance and future-proofness challenges of increasing hardware diversity, by defining a virtual 1024-bits instruction set that consists of simple operators supported by all SIMD dialects; and also, importantly, by scalar code. The interleaved and tuple-reordered layout actually makes scalar decoding faster, extracting more data-parallelism from today's wide-issue CPUs. Importantly, the scalar version can be fully auto-vectorized by modern compilers, eliminating technical debt in software caused by platform-specific SIMD intrinsics. Micro-benchmarks on Intel, AMD, Apple and AWS CPUs show that FastLanes accelerates decoding by factors (decoding >40 values per CPU cycle). FastLanes can make queries faster, as compressing the data reduces bandwidth needs, while decoding is almost free.