To perform hot backup, you put the tablespace in begin backup mode, copy the relevant datafiles to either disk or tape and take the tablespace out of backup mode. When you put the tablespace in begin backup mode, Oracle internally performs following operations to maintain consistency:-

• Relevant datafile header is frozen.
• The datafiles are checkpointed.
• Redo generation algorithm is adjusted.
• LGWR begins logging full images of changed blocks to the redo log.
• The datafiles headers which contain the SCN of the last completed checkpoint are not updated while a tablespace is in backup mode.

In normal operation, redo is limited to the minimal change vector applied to a block; in backup mode, no matter how small the change vector is, the entire block is written out to the redo log. This is because you can get a situation where the process of copying the datafiles and DBWn are working on the same block simultaneously. You find that the rate of redo generation against the objects in the tablespace increases substantially and log switch occurs more frequently that triggers more archivelog generation. To limit the effect of this additional logging, make sure you have placed one tablespace at a time in begin backup mode, and take the tablespace out of backup mode as soon as you have backed it up. This will reduce the number of blocks that may have to be logged to the minimal possible.

As we know that Oracle database blocks are a multiple of O/S blocks. Let’s say your O/S has a default block size of 512 bytes, while Oracle’s default block size is 8192 bytes. This means that O/S filesystem stores data in 512 bytes chunks, while Oracle performs operation in 8192 bytes chunks or multiples thereof. While backing up datafiles from the filesystem, if DBWn are triggered to modify a database block at the same moment, then backup copy contains some O/S blocks from before the DBWn performed the write and some from after. Now your backup copy has got a fractured block (split block) means the head and tail are from two different points in time. By logging the full block image of the changed block to the redo log, Oracle guarantees that in the event of recovery, any fractured block that might be in the backup copy of the datafiles will be resolved by overlaying them with the full image of the blocks from the archivelogs.

Once the backup is completed and tablespace is taken out of backup mode, Oracle advances checkpoint SCN for each file to the latest actual SCN value.

Prerequisites of performing hot backup:-
We need to make sure the following :
* Database is open
* Database should be in archivelog mode
* Backup directory is already created
* Undo tablespace and archivelog destination have enough free space
* Hot backup script is ready to execute

Note: there is a popular misconception regarding placing tablespace in begin backup mode is that DBWn stops writing to datafiles, change vectors which are recorded in redo logs are written back into datafiles when a tablespace is taken out of backup mode. Some DBAs think so ….