Peer to Peer Replication en SQL Server 2008

Les adjunto un excelente artículo sobre PeerToPeer Replication en SQL Server 2005, fue escrito por

Dinesh Asanka

http://www.sql-server-performance.com/articles/dba/peer-to-peer_replication_p3.aspx

Using Peer-To-Peer Transactional Replication

Let's consider a typical online e-commerce application. In order to avoid downtime and reduce the load on any single server, the database for this application is in more than one location. As it is an online e-commerce system, data needs to be changed (through inserts, updates, and deletes) at each location and all the data modifications need to be replicated to the other servers. For example, let's assume there are databases at location A, B, and C.

To accomplish this in SQL Server 2000, we would use merge replication. Location A is the publisher, and B and C are the subscribers.

clip_image001
Figure 1: Merge replication.

The obvious drawback of this method is that it is a single point of failure configuration. If database B is down, the A to C replication link will continue to work. Users who are connected to both servers A and C will not see any adverse effect. However, if database A fails, B and C will be isolated from the system. Modifications will not be visible on other end.

SQL Server 2000 replication uses the publisher and subscriber hierarchy method. Successful operation of this configuration requires that the publisher be present at all times.

With the peer-to-peer replication topology in SQL Server 2005, each node acts as a publisher and as a subscriber. Replication recognizes when changes have occurred on a given node but only allows those changes to cycle through the nodes one at a time.

clip_image002
Figure 2: Peer-to-peer replication.

If one database is down (A, for example), the other databases (B and C) can still replicate. Whenever that database (A) comes back up, it can synchronize with the others (B and C) and get the changes that took place after it went down. This is possible because the databases (A, B, and C) all act as both publisher and subscriber.

How to Configure Peer-to-Peer Replication

Now that we have seen how peer-to-peer transactional replication works, let's look at the configuration options.

First, we need to create a publication. In SQL Server 2005, we'll use the New Publication Wizard.

clip_image003
Figure 3: New Publication Wizard.

As you can see (above), there are four publication types. To configure peer-to-peer transactional replication, we need to select Transactional publication.

After the publication is created, we need to change its properties. In the Publication Properties window (see below), set Allow peer-to-peer subscriptions to True. Please note that once we set this property to "true" we cannot revert to "false" until replication for the publication is dropped.

clip_image004
Figure 4: Publication Properties.

Now we'll go back, right click on the publication, and run the peer-to-peer transactional replication wizard (see below).

With this wizard, we can add more SQL Servers or SQL Server instances to the peer network. You will not be able to add main publication databases or the previously selected database on which we ran the wizard.

clip_image005
Figure 5: Configure Peer-To-Peer Topology Wizard: selecting peers.

After configuring the peer network, we need to specify how we have initialized the databases for the other servers in the peer-to-peer network.

clip_image006
Figure 6: Configure Peer-To-Peer Topology Wizard: initializing the new peers.

The obvious way to initialize the new peer databases is to restore the backups (making sure they are not changed before setting up peer-to-peer replication). Otherwise, we can specify the backup file used. The SQL Server engine will synchronize the databases by comparing the publication database to the backup.

After we provide authentication for the Log Reader and the SQL Agent, the wizard will begin building the peer-to-peer topology. For our example, the process results in the creation of three publications and six subscriptions (see below). For each publication, the wizard will create a distribution database as well.

clip_image007
Figure 7: Configure Peer-To-Peer Topology Wizard: building the topology.

When we have completed the wizard, this is what we'll see in SQL Server 2005 explorer:

clip_image008
Figure 8: The results of running the wizard.

Introducing Another Node

In today's competitive environment, businesses expand widely and frequently. What if we need to add another node? In SQL Server 2005, using peer-to-peer replication, all we need to do is configure the new node as a publisher and a subscriber and attach it to the peer-to-peer network.

Conclusion

Peer-to-peer transactional replication has added a new dimension to SQL Server database replication, giving DBAs more flexibility. As a new feature, it isn't yet widely used. But it will be. So check it out and see how your organization will benefit.

Comments

Anonymous said…
Great break down of how Peer to Peer works - Thank you!. We have been trying to set this up and are having issues. POnce we set up the publication (and distribution) we find that in the publication properties, we do not have an option for 'peer to peer' and are therefore unbale to set to true! Similalry we dont have the option for starting peer to peer replication wizard when right clicking the publication. Any ideas?
Eduardo Castro said…
When you are creating the publication you can choose to use peer to peer replication

Popular posts from this blog

Desempeño de SQL Server 2008 R2 y Max Worker Threads

Cómo identificar consultas más pesadas en SQL Server

SQL Server La longitud de los datos LOB (2200100) que se van a replicar excede el máximo configurado 65536.