I still do not understand why 5+ times ?

Aren't you better off using Snapshots instead ?

And that is *a lot* faster than the whole thing. Of course Snapshots are to...

Simon,

I use BULK_LOGGED before applaying the snapshot and it *DOES* makes a difference, Simple is good too. Full and snapshot is a bad combination for large tables...

Just my $0.02

On way to de-couple it is to actually queue requests on the outside that are read from the inside, applied inside and then replicated to the outside. What that means...

You can set the snapshot to "DROP" objects in the replica before applying it.

In addition that replica should be protected from client writes ( this can be achived using security...

Before doing so I would recommend you review :

Referencing parent nodes in XQuery is EXPENSSIVE

Cheers,

you are missing the Logins..

I think you will be better off programing your custom logic in stored procedures!!!

I am not sure your assumption is correct. Subscribers *can* change the data and the "conflict resolution policy" determines who wins if "both" change the same record.

shrink distribution transaction log.

peer-to-peer does also gives you what you ask. This setup is *very risky* in terms of security. You should try to de-couple the logic of the outside modifications with the...

Those errors occur normally either when someone "modified" the replica and the snapshot does not "truncate or drop" the table at the destination

:hehe::hehe::hehe::hehe::hehe::hehe::hehe:

What authentication method are you using to connect to SQL Server, Windows authentication or SQL Authentication?

\REPLDATA_BK\unc\LNGOKCSQLP003_BANKRUPTCY_BK-SMALLER TABLES A-C\20071203122870is a SNAPSHOT folder. What is failing on that error is the Snapshot Agent.

You should check for the credentials of that snapshot Job.

Secondly the Cleanup Job will log...

BULK INSERT / SELECT INTO would have been twice ( maybe more) as fast and yes recovery model to BULK_LOGGED is mandatory