当前位置: X-MOL 学术J. Netw. Syst. Manag. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
A Workload-Driven Approach for View Selection in Large Dimensional Datasets
Journal of Network and Systems Management ( IF 3.6 ) Pub Date : 2020-03-14 , DOI: 10.1007/s10922-020-09526-z
Leandro Ordonez-Ante , Gregory Van Seghbroeck , Tim Wauters , Bruno Volckaert , Filip De Turck

The information explosion the world has witnessed in the last two decades has forced businesses to adopt a data-driven culture for them to be competitive. These data-driven businesses have access to countless sources of information, and face the challenge of making sense of overwhelming amounts of data in a efficient and reliable manner, which implies the execution of read-intensive operations. In the context of this challenge, a framework for the dynamic read-optimization of large dimensional datasets has been designed, and on top of it a workload-driven mechanism for automatic materialized view selection and creation has been developed. This paper presents an extensive description of this mechanism, along with a proof-of-concept implementation of it and its corresponding performance evaluation. Results show that the proposed mechanism is able to derive a limited but comprehensive set of views leading to a drop in query latency ranging from 80% to 99.99% at the expense of 13% of the disk space used by the base dataset. This way, the devised mechanism enables speeding up query execution by building materialized views that match the actual demand of query workloads.

中文翻译:

大维数据集中视图选择的工作负载驱动方法

过去二十年世界见证的信息爆炸迫使企业采用数据驱动的文化来提高竞争力。这些数据驱动的业务可以访问无数信息源,并面临着以高效可靠的方式理解海量数据的挑战,这意味着执行读取密集型操作。在这一挑战的背景下,已经设计了一个用于大维度数据集的动态读取优化的框架,并在此基础上开发了一种用于自动物化视图选择和创建的工作负载驱动机制。本文对这种机制进行了广泛的描述,以及它的概念验证实现及其相应的性能评估。结果表明,所提出的机制能够派生出有限但全面的视图集,从而以基础数据集使用的磁盘空间的 13% 为代价,将查询延迟降低 80% 至 99.99%。这样,设计的机制可以通过构建与查询工作负载的实际需求相匹配的物化视图来加速查询执行。
更新日期:2020-03-14
down
wechat
bug