About James Sutherland
One of the most effective database optimizations is batch writing. Batch writing is supported by most modern databases and part of the JDBC standard and is supported by most JPA providers.
Normal database access consists of sending each DML (insert, update, delete) statement to the database in a separate database/network access. Each database access has a certain amount of overhead to it, and the database must process each statement independently. Batch writing has two forms, dynamic and parametrized. Parametrized is the most common, and normally provides the best benefit, as dynamic can have parsing issues.
To understand batch writing, you must first understand parametrized SQL. SQL execution is composed of two parts, the parse and the execute. The parse consists of turning the string SQL representation to the database representation. The execute consists of executing the parsed SQL on the database. Databases and JDBC support bind parameters, so the the arguments to the SQL (the data) does not have to be embedded in the SQL. This avoids the cost of converting the data into text, and allows for the same SQL statement to be reused, with multiple executions. This allows for a single parse and multiple executes, aka “parametrized SQL”. Most JDBC DataSource implementations and JPA providers support parametrized SQL and statement caching, this effectively avoids ever having a parse in a running application.
Source : http://www.javacodegeeks.com/2013/09/batch-writing-and-dynamic-vs-parametrized-sql-how-well-does-your-database-perform.html
