Benchmarking learned indexes

Marcus, Ryan; Kipf, Andreas; Renen, Alexander van; Stoian, Mihail; Misra, Sanchit; Kemper, Alfons; Neumann, Thomas; Kraska, Tim

doi:10.14778/3421424.3421425

Cited by 106 publications

(90 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We evaluate all join algorithms with real and synthetic datasets. For real datasets, we use three datasets from the SOSD benchmark [37], where we perform a self-join with each dataset. Any dataset is a list of unsigned 64-bit integer keys.…”

Section: Methodsmentioning

confidence: 99%

“…According to a recent benchmark study [37], both RMI and RadixSpline are considered the most efficient read-only learned indexing structures. Therefore, in this paper, we explore their usage in improving the performance of indexed nested loop joins as described later in Section 3.…”

Section: Cdf-based Partitioning and Learned Indexing Structuresmentioning

confidence: 99%

“…Our experimental evaluation shows that the proposed learned joins for INLJ and SJ provide the best performance against the state-of-the-art competitors in each category with different real (SOSD [37] and TPCH [55]) and synthetic benchmarks. For example, our Gapped RMI INLJ can be 2.7X faster than hash-based INLJ in many scenarios and with different datasets.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

The Case for Learned In-Memory Joins

Sabek¹,

Kraska²

2021

Preprint

Self Cite

View full text Add to dashboard Cite

In-memory join is an essential operator in any database engine. It has been extensively investigated in the database literature. In this paper, we study whether exploiting the CDF-based learned models to boost the join performance is practical or not. To the best of our knowledge, we are the first to fill this gap. We investigate the usage of CDF-based partitioning and learned indexes (e.g., RMI and RadixSpline) in the three join categories; indexed nested loop join (INLJ), sort-based joins (SJ) and hash-based joins (HJ). We proposed new efficient learned variants for the INLJ and SJ categories. In addition, we proposed a reinforcement learning based query optimizer to select the best join algorithm, whether learned or not-learned, for each join query. Our experimental analysis showed that our learned joins variants of INLJ and SJ consistently outperform the state-of-the-art techniques.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Cdf-based Partitioning and Learned Indexing Structuresmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

The Case for Learned In-Memory Joins

Sabek¹,

Kraska²

2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Recently, machine learning (ML) found its way into indexing. While learned indexes [26,24,32,25] may consume significantly less space than traditional indexes such as Btrees, that observation mostly applies to primary indexing where the base data is already sorted. In a secondary indexing setting such as ours, additional permutation vectors have to be maintained which will likely account for most of the space consumption of the learned index.…”

Section: Related Workmentioning

confidence: 99%

Cuckoo index

Kipf¹,

Chromejko

Hall

et al. 2020

Proc. VLDB Endow.

Self Cite

View full text Add to dashboard Cite

In modern data warehousing, data skipping is essential for high query performance. While index structures such as B-trees or hash tables allow for precise pruning, their large storage requirements make them impractical for indexing secondary columns. Therefore, many systems rely on approximate indexes such as min/max sketches (ZoneMaps) or Bloom filters for cost-effective data pruning. For example, Google PowerDrill skips more than 90% of data on average using such indexes. In this paper, we introduce Cuckoo Index (CI), an approximate secondary index structure that represents the many-to-many relationship between keys and data partitions in a highly space-efficient way. At its core, CI associates variable-sized fingerprints in a Cuckoo filter with compressed bitmaps indicating qualifying partitions. With our approach, we target equality predicates in a read-only (immutable) setting and optimize for space efficiency under the premise of practical build and lookup performance. In contrast to per-partition (Bloom) filters, CI produces correct results for lookups with keys that occur in the data. CI allows to control the ratio of false positive partitions for lookups with non-occurring keys. Our experiments with real-world and synthetic data show that CI consumes significantly less space than per-partition filters for the same pruning power for low-to-medium cardinality columns. For high cardinality columns, CI is on par with its baselines.

show abstract

“…For example, such an index could be used for global dictionary encoding [6]. We introduce the RadixStringSpline (RSS), a learned string index consisting of a tree of the learned index structure RadixSpline (RS) [12,11,15]. RS is a learned index that consists of an error-bounded spline which is in turn indexed in a radix lookup table.…”

Section: Introductionmentioning

confidence: 99%

Bounding the Last Mile: Efficient Learned String Indexing

Spector¹,

Kipf²,

Vaidya³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

We introduce the RadixStringSpline (RSS) learned index structure for efficiently indexing strings. RSS is a tree of radix splines each indexing a fixed number of bytes. RSS approaches or exceeds the performance of traditional string indexes while using 7-70× less memory. RSS achieves this by using the minimal string prefix to sufficiently distinguish the data unlike most learned approaches which index the entire string. Additionally, the bounded-error nature of RSS accelerates the last mile search and also enables a memoryefficient hash-table lookup accelerator. We benchmark RSS on several real-world string datasets against ART and HOT. Our experiments suggest this line of research may be promising for future memory-intensive database applications.

show abstract

Benchmarking learned indexes

Cited by 106 publications

References 21 publications

The Case for Learned In-Memory Joins

The Case for Learned In-Memory Joins

Cuckoo index

Bounding the Last Mile: Efficient Learned String Indexing

Contact Info

Product

Resources

About