Hi Sean,

That is a very good question.

You see, when we are using this technique, we have two or three shadows
going at a time.
We prune one off which, once the shadow is created, seems to take only a
minute, and we add another on, but since we already have the backup in
hand, we do not pay attention to the time it takes to create a new
shadow.

When I started playing with shadows (and pestering poor Ann with alot of
questions about changing page sizes and other such nonsense) I found
that they seemed faster (by 6 hours or so in comparison to the backup
process). BUT, please remember, that the only way to take a proper
reading is to have the shadow file on a separite data bus (scsi channel)
and to have the whole storage device dedicated to it. Then to try the
same thing with a backup. And you need to do the comparison without any
others attached to the database.
You see, both user and hardware contention can skew the numbers in any
direction. If you are going to perform tests, you need an environment a
little less chaotic than what we have here.

We use this technique for our test databases when we may need to fork a
database to try a new process or technique while not disturbing the
developements of others in the office.

best regards

Dalton

"Leyne, Sean" wrote:

>
> Dalton,
>
> How does the performance of the create shadow compare to performing a
> full backup of your DB?
>
> Sean