An Evaluation of Cassandra for Hadoop

Abstract

Abstract-In the last decade, the increased use and growth of social media, unconventional web technologies, and mobile applications, have all encouraged development of a new breed of database models. NoSQL data stores target the unstructured data, which by nature is dynamic and a key focus area for "Big Data" research. New generation data can prove costly and unpractical to administer with SQL databases due to lack of structure, high scalability, and elasticity needs. NoSQL data stores such as MongoDB and Cassandra provide a desirable platform for fast and efficient data queries. This leads to increased importance in areas such as cloud applications, e-commerce, social media, bio-informatics, and materials science. In an effort to combine the querying capabilities of conventional database systems and the processing power of the MapReduce model, this paper presents a thorough evaluation of the Cassandra NoSQL database when used in conjunction with the Hadoop MapReduce engine. We characterize the performance for a wide range of representative use cases, and then compare, contrast, and evaluate so that application developers can make informed decisions based upon data size, cluster size, replication factor, and partitioning strategy to meet their performance needs.

A Processing Pipeline for Cassandra Datasets Based on Hadoop Streaming

“…In previous studies [11], [14], we show that in such cases the write location becomes a strong determinant of MapReduce performance. Therefore, in Figure 6 we include two different setups for the Hadoop native and Cassandra consolidation.…”

“…These, observed commonly in MapReduce use cases, are not necessarily considered to be processing heavy workloads and considerable portion of the job time is spent on reading the input. As the output data size is small, the write times have a minimal effect on job performance [11]. In Figure 5 we show the relative performance of running such examples of non-Java executables using MARISSA, Hadoop Streaming and Hadoop-C*.…”

“…These additional steps contribute to the total job time significantly. That is, leaving the data in Cassandra servers and using Hadoop-C* provides better performance even with the considerable overhead observed when reading the input from the Cassandra servers [11]. At four million records, Hadoop-C* is 1.7 times faster than MARISSA while it is 2.6 times faster for 64 million records.…”

“…2) MR2: Write Intensive Applications: Figure 6 shows the performance of running "Reorder" and "Merge" type operations that are classified as write intensive workloads [11] and [14]. In this group of applications, the size of the output is either nearly equal or greater than the input dataset.…”

Big Data: una exploración de investigaciones, tecnologías y casos de aplicación

“…Cassandra ofrece un modelo de datos que cuenta con comodidad para la indexación de columnas, soporte a la desnormalización y materialización a las vistas y un poderoso almacenamiento en caché integrado. Es un sistema de almacenamiento distribuido con un modelo de datos que soporta un control dinámico sobre el diseño y el formato de los datos [30]. Algunos de los principales atributos de Cassandra son:…”