JODA: A Vertically Scalable, Lightweight JSON Processor for Big Data Transformations

We describe the demonstration of JODA (Json On Demand Analytics), an approach to handling large amounts of JSON documents in a vertically scalable manner. With JODA, the user can import, filter, transform, aggregate, group, and export documents with a simple PIG-style query language, offering fast execution speed. This is achieved by utilizing a multithreaded architecture over disjoint, read-only containers of data that are processed in parallel, similar to what RDDs are to Spark. Containers are augmented with auxiliary information like Bloom filters and adaptive indices and all containers are processed in parallel by individual threads. By avoiding locks, latches, and synchronization beyond simple thread pooling, we do not risk contention and therefore maximize resource utilization. The demonstration scenarios aim at engaging visitors with several data analytics tasks around large, real-world datasets that are to be solved with the help of JODA, and further gives insights on system internals and the installation/configuration process.