In a data warehouse, loading/transforming data is a significant process. Do Informix non-logged databases provide significant performance gains in loading and transforming data?
YES! Though your needs may vary, most databases in data warehouses are not logged. There is very little change within the database that needs to be tracked on a transaction-by-transaction basis. Data is loaded once or twice a day and, depending on how big the warehouse is, and based on the archiving scheme used, the load can be backed out and restarted if necessary.
With the databases in a non-logged mode, each row that is added, modified or deleted does not have to be logged through the physical and logical log structures. The overhead this process requires can be substantial depending on the amount of data being moved around. You also do not have to watch out for long transaction conditions in which a single data manipulation command will affect so many rows that the acceptable limit of logical log usage is reached. In non-logged databases, only DDL statements (e.g. create table, drop index statements), not DML statements (insert, update, delete SQL commands), are logged in the logical logs,.
In an OLTP environment, the overhead resulting from writing to the physical and logical logs is acceptable because the number of rows affected by a single transaction is relatively small, so long transaction conditions are usually not a problem. In addition, the environment can not afford to lose a single transaction, so the ability to track changes in the database is critical.
So to save yourself time and grief, run your data warehouse databases in a non-logged mode.