November 18, 2008 at 4:55 am
Hi All,
We have a live replication system already setup, but I am looking at a new system where we can take replication duties away from our main customer server, the idea is to have another box seperate from the customer live server, this box will be the one configured to do the replication (publisher and distributor), this server will in turn get articles from the central customer live server, so in turn it will be replicating data that its currently subscribed to.
I do not know if this configuration will work, I will like to get some information about how the sytem can be configured.
Cheers.
November 18, 2008 at 8:31 am
Hi,
Yes you can; have a look at "Republishing data" in Books Online.
/Elisabeth
elisabeth@sqlserverland.com
MCITP | MCT
http://sqlblog.com/blogs/elisabeth_redei/
http://linkedin.com/in/elisabethredei
November 18, 2008 at 8:46 am
would it not be simpler to just move your distributor to another server and have the distributor runing all of the log reader and distribution agents?
MVDBA
November 18, 2008 at 3:34 pm
michael vessey (11/18/2008)
would it not be simpler to just move your distributor to another server and have the distributor runing all of the log reader and distribution agents?
Hi,
Thanks for your thoughts, its just that this server is quite central to the operations of the company and I am just worried that if anything goes wrong, it can cause a lot of problems, besides I would not want a situation where we might be having performance problems and eyebrows will be raised as to whether replication is the cause of the problem.
November 20, 2008 at 9:42 am
I agree with other posters - sounds like what you need here is a dedicated distributor. Moving distribution responsibilities off your core server will HELP performance because you're transferring the I/O, memory, and CPU overhead of replication to another box.
November 21, 2008 at 2:01 am
kendal.vandyke (11/20/2008)
I agree with other posters - sounds like what you need here is a dedicated distributor. Moving distribution responsibilities off your core server will HELP performance because you're transferring the I/O, memory, and CPU overhead of replication to another box.
Is the distributor what really does the replication work, does the publilsher take any hit in terms of performance in replication ?
November 21, 2008 at 7:43 am
Yes, the distributor is the workhorse for replication. The publisher simply writes entries to the transaction log. The distributor is what holds the distribution database containing all the replicated commands and runs all the agents. Once the log reader agent reads through the transaction log back on the publisher all the work for replication happens on the distributor.
This BOL article explains how transactional replication works and shows the role of the distributor:
Viewing 7 posts - 1 through 6 (of 6 total)
You must be logged in to reply to this topic. Login to reply