Scalable Linear Algebra on a Relational Database System

Luo, Shangyu; Gao, Zekai J.; Gubanov, Michael; Perez, Luis L.; Jermaine, Christopher

doi:10.1145/3277006.3277013

Cited by 24 publications

(22 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Existing works [12,45,50] propose to: (1) Abstract the tensor as a set of tensor blocks; (2) Encode local linear algebra computation logics that manipulate single or a pair of tensor blocks, in user defined functions (UDFs), also called as kernel functions, such as matrix multiplication, matrix addition, etc. ; (3) Apply the relational algebra operators nested with these UDFs for performing linear algebra computations.…”

Section: Background and Related Work 21 ML Model Inferences As Queriesmentioning

confidence: 99%

See 1 more Smart Citation

Serving Deep Learning Models with Deduplication from Relational Databases

Zhou¹,

Chen²,

Das³

et al. 2022

Preprint

View full text Add to dashboard Cite

Serving deep learning models from relational databases brings significant benefits. First, features extracted from databases do not need to be transferred to any decoupled deep learning systems for inferences, and thus the system management overhead can be significantly reduced. Second, in a relational database, data management along the storage hierarchy is fully integrated with query processing, and thus it can continue model serving even if the working set size exceeds the available memory. Applying model deduplication can greatly reduce the storage space, memory footprint, cache misses, and inference latency. However, existing data deduplication techniques are not applicable to the deep learning model serving applications in relational databases. They do not consider the impacts on model inference accuracy as well as the inconsistency between tensor blocks and database pages. This work proposed synergistic storage optimization techniques for duplication detection, page packing, and caching, to enhance database systems for model serving. Evaluation results show that our proposed techniques significantly improved the storage efficiency and the model inference latency, and serving models from relational databases outperformed existing deep learning frameworks when the working set size exceeds available memory.

show abstract

Section: Background and Related Work 21 ML Model Inferences As Queriesmentioning

confidence: 99%

“…Therefore, as illustrated in Fig. 1, a fully-connected feed-forward network (FFNN) can be represented in relational algebra [34,45]. While the experiments in this work (Sec.…”

Section: Background and Related Work 21 ML Model Inferences As Queriesmentioning

confidence: 99%

Serving Deep Learning Models with Deduplication from Relational Databases

Zhou¹,

Chen²,

Das³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Advanced in-database analytics. To accommodate the exponential growth in data science and machine learning applications, a recent line of work [3,14,24,26,39,41,54,77] focuses on supporting advanced analytics queries that involve linear algebra (LA) operators. TCUDB shares the goal of LevelHeaded [3] in identifying the worst-case optimal join (WCOJ) [62] or LaraDB's rule-based translation between relational queries and parallel LA queries, but TCUDB additionally provides the capability of translating (parts of) the query to TCU-accelerated matrix multiplication operator(s) and different sets of opportunities from the orders of magnitude speedup by TCUs in such operations.…”

Section: Related Workmentioning

confidence: 99%

“…Despite being originally designed for AI/ML workloads, tensor processors also hold potential performance improvements for database engines. This is due to both the increasing demand for native support of linear algebra queries (e.g., matrix multiplication itself) in SQL DB engines [3,24,26,39,54] and the observation that a large number of regular query operators can be cast into matrix multiplication. For example, one can show that the most commonly used natural joins [5,20] and group-by aggregates can be encoded as matrix multiplication, which enables TCUs to deliver exceptional performance.…”

Section: Introductionmentioning

confidence: 99%

TCUDB: Accelerating Database with Tensor Processors

Hu¹,

Li²,

Tseng³

2021

Preprint

View full text Add to dashboard Cite

The emergence of novel hardware accelerators has powered the tremendous growth of machine learning in recent years. These accelerators deliver incomparable performance gains in processing high-volume matrix operators, particularly matrix multiplication, a core component of neural network training and inference. In this work, we explored opportunities of accelerating database systems using NVIDIA's Tensor Core Units (TCUs). We present TCUDB, a TCU-accelerated query engine processing a set of query operators including natural joins and group-by aggregates as matrix operators within TCUs. Matrix multiplication was considered inefficient in the past; however, this strategy has remained largely unexplored in conventional GPU-based databases, which primarily rely on vector or scalar processing. We demonstrate the significant performance gain of TCUDB in a range of real-world applications including entity matching, graph query processing, and matrix-based data analytics. TCUDB achieves up to 288× speedup compared to a baseline GPU-based query engine.

show abstract

“…We have briefly discussed enhancements to SimSQL to provide vector and matrix support (see [33] for more detail). Others have also considered to incorporate array data types into relational systems.…”

Section: Related Workmentioning

confidence: 99%

The BUDS Language for Distributed Bayesian Machine Learning

Gao

Luo

Perez

et al. 2017

Proceedings of the 2017 ACM International Conference on Management of Data

Self Cite

View full text Add to dashboard Cite

We describe BUDS, a declarative language for succinctly and simply specifying the implementation of large-scale machine learning algorithms on a distributed computing platform. The types supported in BUDS-vectors, arrays, etc.-are simply logical abstractions useful for programming, and do not correspond to the actual implementation. In fact, BUDS automatically chooses the physical realization of these abstractions in a distributed system, by taking into account the characteristics of the data. Likewise, there are many available implementations of the abstract operations offered by BUDS (matrix multiplies, transposes, Hadamard products, etc.). These are tightly coupled with the physical representation. In BUDS, these implementations are co-optimized along with the representation. All of this allows for the BUDS compiler to automatically perform deep optimizations of the user's program, and automatically generate efficient implementations.

show abstract

Scalable Linear Algebra on a Relational Database System

Cited by 24 publications

References 20 publications

Serving Deep Learning Models with Deduplication from Relational Databases

Serving Deep Learning Models with Deduplication from Relational Databases

TCUDB: Accelerating Database with Tensor Processors

The BUDS Language for Distributed Bayesian Machine Learning

Contact Info

Product

Resources

About