

VerdictDB Documentation

Introduction

VerdictDB is a thin, platform-independent, interactive analytics library that works on top of your existing (or backend) database system (e.g., MySQL, PostgreSQL, Redshift, etc.). For platform-independence, VerdictDB makes all communications with the backend database in SQL. For interactive querying, VerdictDB intelligently infers the query answers based on the results processed on a part of the original data. Those inferred answers are highly accurate estimators of the exact answers. Furthermore, even when only exact answers are needed, VerdictDB can also be useful with its streaming sql engine. The streaming sql engine provides interactive-speed feedbacks in the process of computing exact answers. See this page if you want to know more about VerdictDB's internal mechanism.

Workflow Overview

First, users must create a scramble for their large table. The scramble is just some table in a special format. Once the scramble is created, VerdictDB performs its unique operations to quickly process aggregate queries involving the large table.

Contents

1. Tutorial
   • Quickstart
   • Step-by-step tutorial
     • Simple Application
     • TPC-H Application
       • Setup Databases
       • Load TPC-H Data
       • Example VerdictDB Application with TPC-H
   • Connecting to databases
   • Suppored queries
2. Reference
   • Query Syntax
     • VerdictDB JDBC Properties
     • Simple Querying
     • Creating/Viewing Scrambles
     • Appending Scrambles
     • Dropping Scrambles
     • Select-Query Syntax
     • Stream Querying
   • Setting up TPC-H dataset
   • Javadoc
   • Pydoc
3. How it works
   • Basics
   • Architecture
   • Query Processing

License and Developments

VerdictDB is under the Apache License 2.0; thus, it is completely free for both commercial and non-commercial purposes. VerdictDB is developed by the database group at the University of Michigan, Ann Arbor.