Layering a DBMS on a DHT-Based Storage Engine

Abstract

This article proposes an architecture for integrating a relational database management system (DBMS) with a distributed hash table (DHT).   Systems developed based on this architecture provide functionalities of DHTs, such as scalability, decentralization and fault tolerance, combined with a high level query language provided by a DBMS. The main component of this architecture is the storage engine,  responsible for implementing the interface between an SQL query processor and a DHT. We have implemented a system based on the MySQL DBMS and the Bamboo DHT, and conducted an experimental study
to investigate two issues. The first determines the impact of developing an indexing structure  for processing range queries. The second investigates two approaches for mapping relations to DHT's key-value pairs: vertical and horizontal partition of relations.