Search

Non standard approach to the implementation of relational databases this work implements a database system storing the data by columns instead of rows. It is based on the C Store system design and is limited only to the fundamental solution issues. Two areas are being solved: the initial creation of the database, including import of static database data from another source, and querying. The database does not store the data in tables but in projections groups of columns of the same length. By using this database layout it is possible to evaluate queries faster than with database storing by rows as the database engine does not have to read irrelevant columns. Querying is allowed by providing an execution plan consisting of the C Store query operators that work with columns or projections. Composition of these operators can replace queries in SQL notation. This work sets the rules of behavior of various operators, implements these operators and also mentions the optimization of the execution plan. The optimization also includes a C Store system extension providing direct access to data items, which in some cases radically reduce the amount of IO operations. The practical output of the work is a library that provides a programming interface for database access and manipulation.