We're also looking at the Ontrack software for quick file restores.  I 
got a tip from a reader on this list, of how to increase the mailbox 
backup speeds.  It involved breaking down the usernames into one data 
stream per letter of the alphabet.   The message thread was "Brick-level 
Exchange Backup Performance", back in October of 2003.

The list I initially tried was designed to get as many data streams out 
of the exchange server as possible.  In our environment, this caused a 
lot of problems, and I would get exit status 41 on several streams, 
where it would almost always give exit staus 1 when using one stream.  
It really was faster than one stream, obviously, but with the increased 
errors, and errors other than status 1, I wasn't comfortable with the 
reliability of the backup.  I decided to reduce the number of streams by 
grouping the users into larger groups of the alphabet.  I now have four 
groupings, and my errors have almost disappeared.  What used to take 
almost 6 or 8 hours to do, now takes under 3 hours.  I have also noticed 
that the multiple data streams improve the speed to search out a message 
during recovery.  Our "information store" backup takes just over 90 
minutes, and it's a full backup every night.  That's why we're so 
interested in the Ontrack software.  It's priced based on mail accounts, 
and it works extremely fast! 

Splitting the mailboxes improves speed, but will load up your exchange 
server as it works to deliver the data. 
Four streams for about 200 users seems to be an acceptable load on our 
system.  Our exchange database fluctuates between 50 and 60 gig.



-Jon Bousselot



> What we do to handle this is a little different since brick level 
> backup of exchange is slooooow
>
> We restore the full database from the latest full-backup (we only use 
> full backps) to a server/workstation that has/had exchange installed 
> (but not active) and then we use a toll called Ontrack to pull out the 
> objects we need. Ontrack doesn't cost much and it's a simple and great 
> tool.
>
> /johnny
>
> At 09:07 26-11-2003 +1100, Broun, Bevan wrote:
>
>> on Tue, Nov 25, 2003 at 01:29:24PM -0500, Jim Langston 
>> <jlangston AT northeastmedical DOT org> wrote:
>> >
>> > First question:  Is it normal for the brick level backups to run 
>> this slow?
>> > Second question: If not, what are some of the ways to configure 
>> this to run faster?
>>
>> >From what I can tell from this list, this speed is normal. I am 
>> considering
>> a 2nd exchange server setup just to do database restores to. This would
>> provide practise at restore and it should be possible to extract 
>> mailboxes
>> from the restore server.
>>
>> Im not an exchange person, so I dont know what issues are involved in
>> restoring databases from one exchange server to the other. Any 
>> suggestings
>> from the list would be good.
>>
>> BB
>>
>