Which one of the following is a key factor for preferring $B$-trees to binary search trees for indexing database relations?
- Database relations have a large number of records
- Database relations are sorted on the primary key
- $B$-trees require less memory than binary search trees
- Data transfer from disks is in blocks