Publication: Random testing basado en lógica temporal para Apache Flink
Loading...
Official URL
Full text at PDC
Publication Date
2019
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
Actualmente, existen muy pocas alternativas para probar los sistemas de stream processing, consistiendo la mayoría de ellas en tests de unidad, los cuales no son viables en casos en los que se requiera disponer de una gran cantidad de streams o streams de gran longitud, ya que es necesario definir cada stream de entrada y el esperado de salida. Es por esto que surgió la idea de implementar un programa que pueda aplicar random testing, una técnica de testing basada en propiedades en lugar de en coberturas, a uno de estos sistemas de stream processing, incorporando además propiedades de lógica temporal. Además, ha sido empíricamente demostrado que el random testing funciona igual o incluso mejor que las técnicas de cobertura, lo que inclina la balanza a favor de esta técnica de testing. En este trabajo presentamos una herramienta para realizar testing basado en propiedades para Apache Flink, un sistema de stream processing capaz de procesar datos a tiempo real. Esto significa que Flink trata los datos a medida que van siendo generados y en el momento en que son recibidos. Para desarrollar esta herramienta se ha utilizado el lenguaje Scala y la técnica de random testing combinada con lógica temporal. Ya existe un entorno con la misma filosofía que el aquí presentado dirigido a Spark Streaming, Sscheck. Sin embargo, Spark maneja lotes y no tiene, por tanto, tiempo real. Por ello, el objetivo de este proyecto es implementar ese mismo programa pero adaptado para trabajar con Apache Flink, aunque presenta diferentes problemas, como menor flexibilidad en el tratamiento de los datos.
Nowadays, there are very few alternatives for stream processing systems testing, most of them consisting of unit tests, which are not viable when it is required a large number of streams or very long streams, because it is necessary to define every stream received as input and every stream expected as corresponding output. Because of this situation, the idea of implementing a program that can use random testing, a testing technique based in properties instead of coverage, with one of these processing systems, came up, also adding temporal logic properties. Furthermore, it has been empirically proved that random testing works as well as, or even better than, coverage techniques, something that makes the choice of using this technique even more attractive. In this Master’s Thesis we present a property-based testing tool for Apache Flink, a stream processing system capable of processing data in real time. This means that Flink treats data as it is generated and as it is received. To develop this tool we have used the Scala language and random testing techniques combined with temporal logic. There exists an environment with the same philosophy as the one presented here applied to Spark Streaming, Sscheck. However, Spark handles batches and therefore has no real time. Therefore, the aim of this project is to implement the same program but adapted to Apache Flink. It presents different problems, such as less flexibility in data processing.
Nowadays, there are very few alternatives for stream processing systems testing, most of them consisting of unit tests, which are not viable when it is required a large number of streams or very long streams, because it is necessary to define every stream received as input and every stream expected as corresponding output. Because of this situation, the idea of implementing a program that can use random testing, a testing technique based in properties instead of coverage, with one of these processing systems, came up, also adding temporal logic properties. Furthermore, it has been empirically proved that random testing works as well as, or even better than, coverage techniques, something that makes the choice of using this technique even more attractive. In this Master’s Thesis we present a property-based testing tool for Apache Flink, a stream processing system capable of processing data in real time. This means that Flink treats data as it is generated and as it is received. To develop this tool we have used the Scala language and random testing techniques combined with temporal logic. There exists an environment with the same philosophy as the one presented here applied to Spark Streaming, Sscheck. However, Spark handles batches and therefore has no real time. Therefore, the aim of this project is to implement the same program but adapted to Apache Flink. It presents different problems, such as less flexibility in data processing.
Description
Facultad de Informática, Departamento de Sistemas Informáticos y Computación, Universidad Complutense de Madrid. Curso 2018/2019.
El código de la aplicación está disponible en:
https://github.com/Valcev/TFM