"Phil" <sherrod@newsgroups.nospam > wrote in message
news:uyGmNsUAIHA.5752@TK2MSFTNGP02.phx.gbl...
> Here's out situation:
>
> We have a database with about 50 million records, full text indexing is
> enabled for one column. About 200,000 new records are added every day
> replacing an equal number of records that expire. All updates to the
> database will be directed to the master server database, so we don't have
> to
> worry about merging updates on multiple databases. We are using SQL
> Server
> 2005 with Windows server 2003 on RAID 0 disks.
>
> We need to replicate the database to accomplish two things: (1) continuous
> availability and redundancy for servicing queries at the rate of about
> 37,000 per hour from customers all over the world, (2) distributed
> processing to improve performance. Our applications will distribute the
> queries between the replicated servers. We are just looking for
> replication
> to keep two SQL servers synchronized.
>
> Transactional replication seems to be the best choice, but we are
> concerned
> about an issue: According to the material we've read, a transaction log
> is
> stored by the distributor until a new snapshot is created. First, we
> don't
> understand why the log can't be cleared as transactions are posted to the
> subscriber database. With 200,000 additions and 200,000 deletions per
> day,
> we are concerned with how quickly the transaction would become
> unmanageable.
> Also, we don't like the idea of having to stop updates to create snapshots
> periodically.

>
> What approach would allow replication to run continuously without
> requiring
> snapshots?
Try to initialize your subscriber via a backup.
>
> Would peer-to-peer replication avoid the snapshot problem?