RE: Some Dataguard is good, lots more must be better?

  • From: "Kevin Closson" <kevinc@xxxxxxxxxxxxx>
  • To: <oracle-l@xxxxxxxxxxxxx>
  • Date: Wed, 20 Sep 2006 12:23:53 -0700

>>>and 42 seconds.  If it is sharing a filesystem with a bunch 
>>>of other databases which are running just fine, how do I 
>>>load up the old files for database 42, if the snapshot is at 
>>>the filesystem level?  


  Thanks for the questions. It is an interesting discussion.
Well, the answer is that the technology is called Filesystem
Snapshots (been done before, VRTS etc), but our implemention (
not released yet) supports directory snapshots. And like I said,
potentially hundreds of thousands of them.  These filesystems
can grow to tremendous size (16TB currently, Pbtye next) and
have a CVM under them that grows online as well. So the model
is, create a directory (db_file_create_dest OMF) in a filesystem
called DB42 that contains all database 42's files and snapshot 
it whenever you want. There are any number of other directories 
in the same filesystem--no matter.

  Every snapshot is a new generation of snapshot 0, if you will.
That is, every directory is in essence a level zero snapshot. If
you want to exchange the read/write snapshot from 0 to say, snapshot
number 42, it is a snapshot CLI command. Now, bear in mind that
all other snapshots are children of snapshot 0, so their value
is dimunished once you start snapshoting the new writable 
generation...note, however, that the new level 0 (writable)
can also be snapshotted...many times.

  It is interesting technology, and most importantly applicable
to much more than just databases since 85% of all data is 
unstructured after all.

Other related posts: