H2O persistence framework for column oriented distributed (NoSQL) databases

H2O persistence framework for column oriented distributed (NoSQL) databases

Dino, Kečo

Cloud architectures are most commonly used in cases when large scale data processing is required. Building applications for cloud architectures requires a lot of engineering experience, especially in cases of data persistence. Persistence in cloud architectures is solved using NoSQL database models. In this paper we are working with column oriented NoSQL database model. Main research goal of this paper is building of new persistence framework for column oriented NoSQL databases. H2O (HBase to Object) framework is created to resolve problem of mapping objects into rows in column oriented database and to provide effective mechanisms for data retrieval. Main focus of this framework is to support persistence of domain models presented by standard UML language. Current implementation supports storing content into HBase NoSQL database. Core engine of H2O framework is built on top of XPath standard. All mappings between domain model attributes and columns in row are represented using XPaths. These paths are used to transform object into row and vice versa. H2O framework contains component for integration with Hadoop map reduce processing library to simplify writing of Hadoop map reduce parallel programs. We took two hardware platforms of same price. First platform have HBase 0.90.1 and H2O installed and other have installed Oracle 11g and Hibernate framework. We are comparing performance of these two platforms from aspects of retrieval and persistence of objects. Result of our comparison is that NoSQL model is better from aspects of retrieval by primary key but shows lower performances in save operations. Keywords: NoSQL, persistence, distributed, HBase, Hadoop, mapping, framework, UML, map-reduce

Conference or Workshop Item
PeerReviewed

2012-05-31

1145