An object mapping for the Cassandra distributed database

• Author(s): Gomes, Pedro ; Pereira, José, 1973- ; Oliveira, Rui Carlos Mendes de
• Date: 2011
• Persistent ID: http://hdl.handle.net/1822/15014
• Origin: RepositóriUM - Universidade do Minho
• Subject(s): Non relational databases; Cassandra; Object mapping
• Document Type: Article
• Language: English

Description

New data management systems aimed at huge amounts of data, such as those pioneered by Google, Amazon, and Yahoo, have recently motivated a profound revolution in database management. These systems are characterized by simple data and query models and by their relaxed consistency, that contrasts with the ubiquitous and widely known relational and ACID model. These changes, in combination with the absence of a high level language for query and manipulation, make it harder for developers to port applications or to leverage existing know-how. In fact, the major proponents of such technologies are precisely those companies that can employ highly skilled developers. This paper bridges this gap between the common programmer and the increasingly popular Apache Cassandra distributed database by providing an object mapping for the Java language in the style of object-relational mappers that are commonly used with SQL databases. In detail, we describe how the object-relational paradigm can be mapped to the new data model and experimentally evaluate the result with a combination of the standard TPC-W benchmark and a representative application from the telecom industry.