Big transactions cause the Tlog to grow.
When you do a clean of a large table, the delete statement can cause a very long transaction
Sometimes it's necessary to prevent this.
So we divide one big delete into several little ones.
A table with call centre data gets +/- 1-mlj records a day. Every day we run a script to delete records with date less than one month ago
To prevent big transactions we do a commit every 1000 records. This is slower we know that but it keeps the Tlog small.
For most DBAs, normalization is an understood concept, a bread and butter bit of knowledge. However, it is not at all unusual to review a database design by a development group for an OLTP (OnLine Transaction Processing) environment and find that the schema chosen is anything but properly normalized. This article by Brian Kelley will give you the core knowledge to data model.