Bob Murdoch wrote:

> I think we have a fairly solid routine, where a scheduled job runs
> every night, performing a backup, restore to a temporary folder, and
> validation of the newly restored database occurs. The newly restored
> database is kept for seven days, so that should we find a problem in
> the production database we can hopefully find a solution within one of
> the seven 'offline' copies.
>
> The backup takes about an hour and a half, the restore takes six
> hours, and the validation about 45 minutes. Although these procedures
> do not require the database to be taken offline, and we do have about
> an eight hour window of little to no usage, I worry that as the size
> of this db grows it will exceed this window and start to have an
> impact on performance of the early-morning users.
>
> I'm wondering how the 'big boys' handle their 100GB - Terabyte
> databases? Are they throwing more hardware at the problem as the size
> increases? Is there something I am missing?

I use a separate computer as backup server. It is a machine with weaker
hardware, but good for backup solution. It has full 24hrs to restore the
databases that are backued-up at primary server once a day. If anything
should happen, we can just plug main server out of network and change
the IP address of backup server (~1 minute downtime). I guess you could
use the same approach as well, if cost of having another machine isn't
too much.

HTH

--
Milan Babuskov
http://fbexport.sourceforge.net
http://www.flamerobin.org