We demonstrate Tensor Query Processor (TQP): a query processor that automatically compiles relational operators into tensor programs. By leveraging tensor runtimes such as PyTorch, TQP is able to: (1) integrate with ML tools (e.g., Pandas for data ingestion, Tensorboard for visualization); (2) target different hardware (e.g., CPU, GPU) and software (e.g., browser) backends; and (3) end-to-end accelerate queries containing both relational and ML operators. TQP is generic enough to supports the TPC-H benchmark, and it provides performance that are comparable to, and often better than, that of specialized CPU and GPU query processors.

Deep Learning (DL), especially Graph Convolutional Networks (GCNs) have revolutionized several domains and applications dealing with unstructured data with non-euclidean and graphical relationships. Constructing large-scale Deep GCNs, however, are bottlenecked by glaring systems issues due to memory blow-ups, runtime slowdowns with random access, and I/O costs. This research abstract identifies various systems and scalability issues and proposes a novel system called GraphGem to handle GCN-centric DL tasks end-to-end. GraphGem tackles the bottlenecks by elevating entire GCN workloads for convenient input declarations by the user, and is inspired by lessons from the databases and machine learning systems worlds. This abstract also highlights the bigger picture of the potential research impact alongside tacking systems constraints and what it may mean for data science and deep learning practitioners going forward.