That was exactly the point. With the information that has been given, you
can put it wherever you want to put it. Wherever the dirstibutor is located
"Patrick Ikhifa" <patisi@pacbell.net> wrote in message
news:u3J8qh2DGHA.3828@TK2MSFTNGP14.phx.gbl...
> Uhmm, I am not so sure I agree with you Mike. Yes there is processing over
> head on what ever server you choose to be the distribution server, however
> generally the norm seems to be the Distribution is usually on the same
> server as the publsiher. Without knowing more about the environment at the
> publisher, it is difficult to give a more informed answer to the question.
>
> If as he says he is looking to implement Transactional Replication, I
> would
> assume that the source server is pretty busy and would not want to be
> burdened with handling the addtional overhaed of a log reader role as
> well
> as a Distribution Database work load. I would err on the side of putting
> the
> distribution db on a different server, and the subscribing server would
> not
> be my first choice. Recall the writer says the objective is a standby
> server, which impleas and I assume that this standby server should be
> prepared and could be called upon to carry the production lload in case of
> some form of service iterruption to the primary production server.
>
> What then happens in a scenario like this? Now we have the "new primary"
> server also hosting the distribution DB, something we do not have in
> production. Not so sure about your comment "If it is on the subscriber,
> you will be adding processing overhead on that server. Either way, you
> have
> overhead. Since you are processing changes on both sides, it really
> doesn't
> matter." I cannot speak much on Mirroring as I am still testing it with
> production volume in our labs here.
> Finally, I think my answer now is we need more information like the
> transaction volume and whether there is the option of a separate
> Distribution server, Synch with backups, High speed interconnect between
> publisher and distributor etc. Happy New Year All.
>
>
> "Michael Hotek" <mike@solidqualitylearning.com> wrote in message
> news:eejgLIWDGHA.528@TK2MSFTNGP09.phx.gbl...
>> You can place the distribution database anywhere that you want to. If it
> is
>> on the publisher, you will be adding processing overhead on that server.
> If
>> it is on the subscriber, you will be adding processing overhead on that
>> server. Either way, you have overhead. Since you are processing changes
> on
>> both sides, it really doesn't matter. One very major point, there is no
>> such thing as instantaneous. The replication engine will only propagate
>> a
>> change when it has been committed. This means there is a latency that
> will
>> always exist before the servers. Ifyou need instantaneous, then you need
> to
>> upgrade to SQL Server 2005 and implement database mirroring in which case
>> you can only process changes on 1 of the machines.
>>
>> --
>> Mike
>> Mentor
>> Solid Quality Learning
>>
http://www.solidqualitylearning.com >>
>>
>> "SSM" <SSM@discussions.microsoft.com> wrote in message
>> news:6F24F043-32F0-4D8B-A6FD-C5032C2A8EDC@microsoft.com...
>> > My goal is to create a warm standby using transaction replication with
>> > instantaneous updating to the subscriber. I only have two servers.
>> > Should I
>> > place the distribution database on the same server as the publisher or
> the
>> > subscriber? Is it even possible to place the distribution database on
> the
>> > subscriber server?
>> >
>> > Thanks.
>>
>>
>
>