From Distributed Database to Cloud Database: A Design Perspective

摘要

From early 1980s until middle 1990s, over fifteen years, a lot of work done on designing distributed databases for a given set of queries. There were two problems that were addressed (i) fragmentation: reducing irrelevant data accessed by a query, and (ii) allocation: reducing the amount of data transferred between sites for executing a query. Many algorithms were designed and evaluated for fragmentation and allocation. In particular, queries change over time, methodology, algorithms, and policies for when and how to redesign a distributed database was proposed. In the last decade, with the advent of modem database systems supported by columnar stores and Hadoop like cloud database systems, the users have been following ad-hoc or pre-determined database design methodologies. In this work, we present backgrounds of distributed database design and cloud database systems, and propose possible solutions for cloud database design based on fragmentation and allocation solutions.