Towards a Reference Architecture with Modular Design for Large-scale Genotyping and Phenotyping Data Analysis: A Case Study with Image Data

With the rapid advancement of computing technologies, various scientific research communities have been extensively using cloud-based software tools or applications. Cloud-based applications allow users to access software applications from web browsers while relieving them from the installation of any software applications in their desktop environment. For example, Galaxy, GenAP, and iPlant Colaborative are popular cloud-based systems for scientific workflow analysis in the domain of plant Genotyping and Phenotyping. These systems are being used for conducting research, devising new techniques, and sharing the computer assisted analysis results among collaborators. Researchers need to integrate their new workflows/pipelines, tools or techniques with the base system over time. Moreover, large scale data need to be processed within the time-line for more effective analysis. Recently, Big Data technologies are emerging for facilitating large scale data processing with commodity hardware. Among the above-mentioned systems, GenAp is utilizing the Big Data technologies for specific cases only. The structure of such a cloud-based system is highly variable and complex in nature. Software architects and developers need to consider totally different properties and challenges during the development and maintenance phases compared to the traditional business/service oriented systems. Recent studies report that software engineers and data engineers confront challenges to develop analytic tools for supporting large scale and heterogeneous data analysis. Unfortunately, less focus has been given by the software researchers to devise a well-defined methodology and frameworks for flexible design of a cloud system for the Genotyping and Phenotyping domain. To that end, more effective design methodologies and frameworks are an urgent need for cloud based Genotyping and Phenotyping analysis system development that also supports large scale data processing. In our thesis, we conduct a few studies in order to devise a stable reference architecture and modularity model for the software developers and data engineers in the domain of Genotyping and Phenotyping. In the first study, we analyze the architectural changes of existing candidate systems to find out the stability issues. Then, we extract architectural patterns of the candidate systems and propose a conceptual reference architectural model. Finally, we present a case study on the modularity of computation-intensive tasks as an extension of the data-centric development. We show that the data-centric modularity model is at the core of the flexible development of a Genotyping and Phenotyping analysis system. Our proposed model and case study with thousands of images provide a useful knowledge-base for software researchers, developers, and data engineers for cloud based Genotyping and Phenotyping analysis system development.