Getting Code Near the Data: A Study of Generating Customized Data Intensive Scientific Workflows with Domain Specific Language

Abstract

<p>The amount of data produced in modern biological experiments such as Nuclear Magnetic Resonance (NMR) analysis far exceeds the processing capability of a single machine. The present state-of-the-art is taking the "data to code", the philosophy followed by many ofthe current service oriented workflow systems. However this is not feasible in some cases such as NMR data analysis, primarily due to the large scale ofdata.</p><p>The objective ofthis research is to bring "code to data", preferred in the cases when the data is extremely large. We present a DSL based approach to develop customized data intensive scientific workjlows capable of running on Hadoop clusters. Our DSL has features to facilitate autogeneration of a Web service front end. These services can be used along with existing service oriented workflow systems. Biologists can use our approach either to implement complete workjlows or expose mini workjlows as services, all without any knowledge of the underlying complications ofthe Cloud environment.</p><p>This poster was presented at the 2nd IEEE International Conference on Cloud Computing Technology and Science, Indianapolis, November 30 - December 3, 2010.</p>