CSE512 Projects (Winter 2014)

QuiViz: Visualizing Physical Query Execution in a Relational Big Data Management System

Umar Javed, Thierry Moreau, Dominik Moritz, Adriana Szekeres

Exploring a JOIN query.

QuiViz is a visualization tool that helps database developers explore and understand query execution and data movement in a distributed database management system. QuiViz provides quick insight into common problems such as data skew or performance bottleneck by leveraging visualization techniques to present: (1) data flow between query operators and between workers, (2) query execution and operator dependencies, (3) cluster utilization, (4) network utilization.

In particular, QuiViz is built to inspect query execution in Myria, a distributed big data management system currently being developed in the UW CSE database group. Myria aims towards building a distributed database platform to provide \emph{big data management and analytics as a service} primarily for scientific applications. The proposed visualizations can easily be applied to other DDBMS as well (e.g. Spark, Hadoop).

Software

Try QuiViz online here. This is running on Google App Engine and uses a small Myria cluster with 8 workers that is running at vega.cs.washington.edu:8777.

Team Responsibilities

Our team of four is led by Dominik, who started working on a query visualization framework for Myria last quarter. The responsibilities for each of the group members are listed below:

Materials

Code

Since it is necessary to run Myria and Myria web to be able to see the visualization, we could not put the code into the repository. Adding the code as submodules was not possible because GitHub could not build the page then. Please see our code at https://github.com/uwescience/myria and https://github.com/uwescience/myria-web. There are running instructions for Myria and Myria web. However, the setup is very complex at the moment.