A New Software Architecture Style for Hadoop Systems

Abstract

Big data provides a challenging environment to store, process and analyze large scale data. Various requirements that impact the architectural solution of big data include the source of data and its involved features such as volume, velocity, variety and the type of data. While processing big data, other challenges raised related to scalability, availability, integrity, concurrency, parallelism and performance. Because of all these features and requirements, building a suitable big data solution needs to consider the architectural solutions to satisfy those different elements of the system. Different software architectural styles exist. The aim of this paper is to provide an overview of the important commonly used architecture styles for building big data systems, compare their benefits, performance and main components. And mostly cover how each style can help in resolving the challenges required by developing big data software systems. The goal is to identify and discuss software architectural issues imposed by Hadoop that impact its performance. A new hybrid architecture system to achieve the requirement for Hadoop is proposed and all possible challenges the Hybrid system could face are investigated.