On Thu, 1 Oct 1998 10:33:38 -0700 (PDT), David Lang
<dlang@diginsite.com> said:
> A properly implemented snapshot system (such as network appliance almost
> is) not only saves the inode info at the time of the snapshot, but also
> saves the contents of blocks that are changed after the snapshot.
> The end result is that storage costs are relitivly small for files that
> are appended (a copy of the inode entries and a copy of the last block of
> the file). for database files taht are truly random access storage costs
> can go through the roof.
I'm not sure what you are trying to say. I do know exactly how
snapshots behave. For database files which are random access, the
cost of a snaphot is purely a function of how much of the database is
modified between snapshots: whether that access is random or
sequential is irrelevant (except that granularity issues are more
pronounced in the random access case).
But the upper limit on the cost of snapshots is still bounded by the
cost of "archiving": keeping separate copies of the file.
Databases are in fact one area where snapshots are extremely useful.
It can be very convenient to snapshot a database before a large batch
run on it and to be able to restore from the snapshot if the batch job
fails for any reason.
--Stephen
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/