Testing Data Transformations in MapReduce Programs

MapReduce is a parallel data processing paradigm oriented to process large volumes of information in data-intensive applications, such as Big Data environments. A characteristic of these applications is that they can have different data sources and data formats. For these reasons, the inputs could c...

Full description

Bibliographic Details
Authors: Morán Barbón, Jesús|||0000-0002-7544-3901, Riva Álvarez, Claudio A. de la|||0000-0001-5592-9683, Tuya González, Pablo Javier|||0000-0002-1091-934X
Format: book part
Publication Date:2015
Country:España
Institution:Universidad de Oviedo (UNIOVI)
Repository:RUO. Repositorio Institucional de la Universidad de Oviedo
Language:English
OAI Identifier:oai:digibuo.uniovi.es:10651/34513
Online Access:http://hdl.handle.net/10651/34513
https://dx.doi.org/10.1145/2804322.2804326
Access Level:Open access
Description
Summary:MapReduce is a parallel data processing paradigm oriented to process large volumes of information in data-intensive applications, such as Big Data environments. A characteristic of these applications is that they can have different data sources and data formats. For these reasons, the inputs could contain some poor quality data that could produce a failure if the program functionality does not handle properly the variety of input data. The output of these programs is obtained from a number of input transformations that represent the program logic. This paper proposes the testing technique called MRFlow that is based on data flow test criteria and oriented to transformations analysis between the input and the output in order to detect defects in MapReduce programs. MRFlow is applied over some MapReduce programs and detects several defects