Efficient Lightweight Compression Alongside Fast Scans

Polychroniou, Orestis; Ross, Kenneth A.

doi:10.1145/2771937.2771943

Cited by 23 publications

(16 citation statements)

References 25 publications

(21 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Let us now turn our attention to data-parallel execution using SIMD operations. There has been extensive research investigating SIMD for database operations [51,50,36,37,38,35,46,44]. It is not surprising that this research generally assumes a vectorized execution model.…”

Section: Data-parallel Execution (Simd)mentioning

confidence: 99%

Everything you always wanted to know about compiled and vectorized queries but were afraid to ask

KerstenTimo¹,

LeisViktor²,

KemperAlfons³

et al. 2018

Proc. VLDB Endow.

View full text Add to dashboard Cite

The query engines of most modern database systems are either based on vectorization or data-centric code generation. These two state-of-the-art query processing paradigms are fundamentally different in terms of system structure and query execution code. Both paradigms were used to build fast systems. However, until today it is not clear which paradigm yields faster query execution, as many implementation-specific choices obstruct a direct comparison of architectures. In this paper, we experimentally compare the two models by implementing both within the same test system. This allows us to use for both models the same query processing algorithms, the same data structures, and the same parallelization framework to ultimately create an apples-to-apples comparison. We find that both are efficient, but have different strengths and weaknesses. Vectorization is better at hiding cache miss latency, whereas data-centric compilation requires fewer CPU instructions, which benefits cacheresident workloads. Besides raw, single-threaded performance, we also investigate SIMD as well as multi-core parallelization and different hardware architectures. Finally, we analyze qualitative differences as a guide for system architects.

show abstract

Section: Data-parallel Execution (Simd)mentioning

confidence: 99%

Everything you always wanted to know about compiled and vectorized queries but were afraid to ask

KerstenTimo¹,

LeisViktor²,

KemperAlfons³

et al. 2018

Proc. VLDB Endow.

View full text Add to dashboard Cite

show abstract

“…We also note that recent work on advanced bit-packing schemes [22,13] and their SIMD implementation [27] focus on the benefits of early filtering and either ignore the high per-tuple cost of bit-unpacking, or just position these bit-packed formats as a secondary storage structure. Our choice for byte-aligned storage mostly avoids this penalty and makes early filtering beneficial in a broader range of query workloads (see Section 5.4).…”

Section: Related Workmentioning

confidence: 99%

“…Vectorization enables the use of SIMD and facilitates algorithms that access data belonging to multiple tuples in parallel. In the database context, SIMD instructions have been used for example, to speed up selection scans [37,35,26], for bit unpacking [35,27], bulk loading [24], sorting [6], and breadth-first search on graphs [32].…”

Section: Related Workmentioning

confidence: 99%

“…The benchmark is conducted on a Haswell i5-4670T CPU with the AVX2 instruction set. The codebase that we used for horizontal bit-packing is the highly SIMD-optimized implementation of [27]. The only bit-packing-related extension is a function that allows us to unpack a single tuple at a given position.…”

Section: Advantages Of Byte-addressabilitymentioning

confidence: 99%

“…While many OLAP systems employ bit-packing techniques to achieve higher compression ratios, we refrain from doing so. Our experiments in Section 5.4 show that even with recent SIMD algorithms [27] for bit-packed data, scans only work very fast if their early filtering is either so selective that there are no qualifying tuples and entire blocks can be skipped, or all tuples in a block qualify. In case early filtering produces a sparse set of qualifying tuple positions, the cost of bit-unpacking the scan columns dwarfs the performance gains made by these recent techniques in early filtering.…”

Section: Introductionmentioning

confidence: 97%

See 2 more Smart Citations

Data Blocks

Lang

Mühlbauer

Funke³

et al. 2016

Proceedings of the 2016 International Conference on Management of Data

102

View full text Add to dashboard Cite

This work aims at reducing the main-memory footprint in high performance hybrid OLTP & OLAP databases, while retaining high query performance and transactional throughput. For this purpose, an innovative compressed columnar storage format for cold data, called Data Block s is introduced. Data Blocks further incorporate a new lightweight index structure called Positional SMA that narrows scan ranges within Data Blocks even if the entire block cannot be ruled out. To achieve highest OLTP performance, the compression schemes of Data Blocks are very lightweight , such that OLTP transactions can still quickly access individual tuples. This sets our storage scheme apart from those used in specialized analytical databases where data must usually be bit-unpacked. Up to now, high-performance analytical systems use either vectorized query execution or "just-in-time" (JIT) query compilation. The fine-grained adaptivity of Data Blocks necessitates the integration of the best features of each approach by an interpreted vectorized scan subsystem feeding into JIT-compiled query pipelines. Experimental evaluation of HyPer, our full-fledged hybrid OLTP & OLAP database system, shows that Data Blocks accelerate performance on a variety of query workloads while retaining high transaction throughput.

show abstract

Make the most out of your SIMD investments: counter control flow divergence in compiled query pipelines

et al. 2019

View full text Add to dashboard Cite

Increasing single instruction multiple data (SIMD) capabilities in modern hardware allows for the compilation of dataparallel query pipelines. This means GPU-alike challenges arise: control flow divergence causes the underutilization of vector-processing units. In this paper, we present efficient algorithms for the AVX-512 architecture to address this issue. These algorithms allow for the fine-grained assignment of new tuples to idle SIMD lanes. Furthermore, we present strategies for their integration with compiled query pipelines so that tuples are never evicted from registers. We evaluate our approach with three query types: (i) a table scan query based on TPC-H Query 1, that performs up to 34% faster when addressing underutilization, (ii) a hashjoin query, where we observe up to 25% higher performance, and (iii) an approximate geospatial join query, which shows performance improvements of up to 30%.

show abstract

Efficient Lightweight Compression Alongside Fast Scans

Cited by 23 publications

References 25 publications

Everything you always wanted to know about compiled and vectorized queries but were afraid to ask

Everything you always wanted to know about compiled and vectorized queries but were afraid to ask

Data Blocks

Make the most out of your SIMD investments: counter control flow divergence in compiled query pipelines

Contact Info

Product

Resources

About