| Multiplexing datafiles in rman [message #269106] |
Thu, 20 September 2007 13:01  |
sroy
Messages: 38
Registered: July 2005
|
Member |
|
|
|
I don't want all the datafiles belonging to one tablespace go into one backupset during the backup. With the parallelsim=4, should I use filesperset=1 to control one datafile per channel? The reason is during the recovery, rman will be able to pull the datafiles from the 4 channels at the same time, thus reducing the overall recovery time.
|
|
|
|
| Re: Multiplexing datafiles in rman [message #269349 is a reply to message #269106] |
Fri, 21 September 2007 15:17  |
sroy
Messages: 38
Registered: July 2005
|
Member |
|
|
|
I think I know the answer. The default value for filesperset parameter is 64, which means 64 datafiles will be grouped together into one backup set. Suppose we have 100 datafiles and decide to set filesperset=4 in the backup command. The result is several datafiles (say 16) belonging to the same tablespace may be contained in more than one backup sets. With parallelism=4, the specific tablespace can be recovered faster than that with the default filesperset value.
|
|
|
|
|
|
| Re: Multiplexing datafiles in rman [message #271940 is a reply to message #269106] |
Wed, 03 October 2007 07:10 |
kudur_kv
Messages: 75
Registered: February 2005
|
Member |
|
|
By What I know,
By degree of parallelism you mean to say the number of channels allocated at the time of backup. You can allocate multiple channels for backing up the database/tablespace quickly.
You can also specify the number of datafiles that need to be bundled in to a backup set. if you mention the 'filesperset' parameter, the number of files you have mentioned will be bundled in to one set.
The number of files in a backup set can also be determined by the 'setsize' parameter as well. For example if you mention a set size of 4 GB, RMAN will bundle as many datafiles as it can within the backup set. The selection of the datafiles will always be random when multiple backup sets are being created.
KV
|
|
|
|
| Re: Multiplexing datafiles in rman [message #272022 is a reply to message #269106] |
Wed, 03 October 2007 12:18 |
sroy
Messages: 38
Registered: July 2005
|
Member |
|
|
Correct. According to the document, the default number of files in each backup set is 4 for datafiles and 16 for archived logs. When we specify MAXSETSIZE, RMAN attempts to limit the size in bytes of the backup sets according to the MAXSETSIZE parameter.
Our objective is spread out the datafiles evenly (preferably one datafile for backup set), no matter how big the datafiles are, and then when it's the time for recovery, we can pull the datafiles faster from each backup set using parallelism. I tested on this scenarios and found it to be the case.
|
|
|
|
] 
Blog
Search
Help
Members
Register
Login
Home
