On the necessity of explicit cross-layer data formats in near-data processing systems

IF 1.5 4区计算机科学 Q3 COMPUTER SCIENCE, INFORMATION SYSTEMS

Distributed and Parallel Databases Pub Date : 2020-04-01 DOI:10.1109/ICDEW49219.2020.00009

Tobias Vinçon, Arthur Bernhardt, Lukas Weber, A. Koch, Ilia Petrov

{"title":"On the necessity of explicit cross-layer data formats in near-data processing systems","authors":"Tobias Vinçon, Arthur Bernhardt, Lukas Weber, A. Koch, Ilia Petrov","doi":"10.1109/ICDEW49219.2020.00009","DOIUrl":null,"url":null,"abstract":"Massive data transfers in modern data-intensive systems resulting from low data-locality and data-to-code system design hurt their performance and scalability. Near-Data processing (NDP) and a shift to code-to-data designs may represent a viable solution as packaging combinations of storage and compute elements on the same device has become feasible. The shift towards NDP system architectures calls for revision of established principles. Abstractions such as data formats and layouts typically spread multiple layers in traditional DBMS, the way they are processed is encapsulated within these layers of abstraction. The NDP-style processing requires an explicit definition of cross-layer data formats and accessors to ensure in-situ executions optimally utilizing the properties of the underlying NDP storage and compute elements. In this paper, we make the case for such data format definitions and investigate the performance benefits under RocksDB and the COSMOS hardware platform.","PeriodicalId":50568,"journal":{"name":"Distributed and Parallel Databases","volume":"40 1","pages":"27-45"},"PeriodicalIF":1.5000,"publicationDate":"2020-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1109/ICDEW49219.2020.00009","citationCount":"6","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Distributed and Parallel Databases","FirstCategoryId":"94","ListUrlMain":"https://doi.org/10.1109/ICDEW49219.2020.00009","RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"COMPUTER SCIENCE, INFORMATION SYSTEMS","Score":null,"Total":0}

引用次数: 6

Abstract

Massive data transfers in modern data-intensive systems resulting from low data-locality and data-to-code system design hurt their performance and scalability. Near-Data processing (NDP) and a shift to code-to-data designs may represent a viable solution as packaging combinations of storage and compute elements on the same device has become feasible. The shift towards NDP system architectures calls for revision of established principles. Abstractions such as data formats and layouts typically spread multiple layers in traditional DBMS, the way they are processed is encapsulated within these layers of abstraction. The NDP-style processing requires an explicit definition of cross-layer data formats and accessors to ensure in-situ executions optimally utilizing the properties of the underlying NDP storage and compute elements. In this paper, we make the case for such data format definitions and investigate the performance benefits under RocksDB and the COSMOS hardware platform.

查看原文本刊更多论文

近数据处理系统中显式跨层数据格式的必要性

在现代数据密集型系统中，由于低数据局部性和数据到代码的系统设计导致大量数据传输损害了系统的性能和可扩展性。近数据处理(NDP)和向代码到数据设计的转变可能是一种可行的解决方案，因为在同一设备上封装存储和计算元素的组合已经变得可行。向新发展方案系统架构的转变要求修订既定原则。在传统的DBMS中，诸如数据格式和布局之类的抽象通常分布在多个层，它们的处理方式被封装在这些抽象层中。NDP风格的处理需要明确定义跨层数据格式和访问器，以确保最佳地利用底层NDP存储和计算元素的属性进行原位执行。在本文中，我们对这样的数据格式定义进行了说明，并研究了在RocksDB和COSMOS硬件平台下的性能优势。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Distributed and Parallel Databases 工程技术-计算机：理论方法

CiteScore

3.50

自引率

0.00%

发文量

审稿时长

>12 weeks

期刊介绍： Distributed and Parallel Databases publishes papers in all the traditional as well as most emerging areas of database research, including: Availability and reliability; Benchmarking and performance evaluation, and tuning; Big Data Storage and Processing; Cloud Computing and Database-as-a-Service; Crowdsourcing; Data curation, annotation and provenance; Data integration, metadata Management, and interoperability; Data models, semantics, query languages; Data mining and knowledge discovery; Data privacy, security, trust; Data provenance, workflows, Scientific Data Management; Data visualization and interactive data exploration; Data warehousing, OLAP, Analytics; Graph data management, RDF, social networks; Information Extraction and Data Cleaning; Middleware and Workflow Management; Modern Hardware and In-Memory Database Systems; Query Processing and Optimization; Semantic Web and open data; Social Networks; Storage, indexing, and physical database design; Streams, sensor networks, and complex event processing; Strings, Texts, and Keyword Search; Spatial, temporal, and spatio-temporal databases; Transaction processing; Uncertain, probabilistic, and approximate databases.