Stormgen - A Domain specific Language to create ad-hoc storm topologies

Abstract

Large-scale distributed data processing has gained significant momentum in research in the past decade. With the introduction of MapReduce, many frameworks have been developed that either implement MapReduce or provide additional functionalities useful in a larger domain. While the framework introduced in the MapReduce paper performs batch-processing of data, Apache Storm performs real-time computation on data. Storm does this with the help of Topologies, and the constituents of the Topology are developed using General-purpose Programming Languages (GPL). A Domain-specific Language (DSL) can provide a higher level of abstraction over GPLs and model the specialized features of a particular domain in a better way. In this paper, we propose the development of Storm Topology generator (Stormgen), a DSL for Storm Topology development, and show how the specifications of this DSL can be utilized during the code generation of exact Storm Topology components in Java. The parser and code generator for Stormgen's syntax are developed using the Eclipse Modelling Framework. The practical use of Stormgen is illustrated with a case study which considers the modelling of a Topology for the Word Count application. � 2014, IEEE.