I have an 18.6G database running on a Windows 2003 server with
hyperthreading-enabled dual Xeon processors, 1GB ram, a raid 5 array,
and using FB 1.5 CS. We do a backup and restore every night, saving
the last seven restored databases just-in-case. One problem that has
plagued us is the amount of time that this backup/restore cycle takes.
We see elapsed times anywhere from 8 hours on a weekend to 10+ on some
weekdays where usage of the production db starts hot and heavy first
thing in the morning.

Our goal has been to reduce the elapsed time of this cycle to less
than eight hours. This allows us to detect any problems with the
database early, and take corrective action before the users arrive.

Some recent comments on this list regarding the speed of local
connections on Linux caused me to experiment with using GBAK on
Windows with the embedded client rather than using localhost. I'm
happy to say that the time for a restore has dropped to just over 5.25
hours.

The only caution I would express is using this technique on a single
processor server. We tested this procedure on a server running FB 1.5
SS. It caused all other processes to respond sluggishly, and
especially affected access to the production copy of the database on
that server.


Bob M..