Peer-to-Peer replication in SQL server '08

BANGALORE, INDIA: With enterprises growing larger each day, it has become customary for most of them to distribute their databases across locations. For instance, a national level company might wish to separate their sales database by state so that each regional sales office can manage their transactions and updates.

However, one of the problems that occur in such a scenario is when all other locations need to see some of these changes and be able to send changes they themselves make across. Till now, replication used to solve some of these issues by using a Publisher-Subscriber model to allow a particular server to be a publisher of some data and other nodes to be subscribers to this data and be able to pull it in.

For instance, master table data could be managed at the central location and all other locations would be able to see changes that occurred there. There were ways to send data 'back' to the publisher as well-but were prone with difficulties such as conflicts arising from same data updates etc.

SQL Server 2008 introduces a new type of data replication called Peer-to-Peer replication. This is actually a subset of Transactional replication, but takes care of the issues that this faced in two-way data updates. P2P replication can work over many nodes on the network and pass data from any node to any other nodes as well.

Refer to the architecture diagram to see a 4-way P2P replication topology. You can scale up the topology (dependent of course on network latency and availability amongst other things) to 15--20 nodes per replication group. Scaling out further can involve creating mesh-like environment of many P2P setups connected to each other.

So how do you setup P2P replication in SQL 2008? For this, you need to have the appropriate access rights to each of the servers that make up the P2P set.

You also need the replication component setup in all the nodes. Once that is in place, you need to open up SQL Server Mgmt Studio (SSMS) and connect to each of the database engine servers. Since this is P2P replication, you can start the configuration from any of the nodes unlike in any of the other replication types.

Select a node and expand the tree in Object Explorer to replication. Right click this and select Configure Distribution. A wizard pops up that walks you through the initial instantiation of the replication engine. For a simple setup, the defaults will be fine. Perform these steps in each node you want to add to the Peer list.

Once this is complete, open Replication | Local Publications. Right click this and select New Publication and follow the wizard. Select a database that you wish to replicate first. In the next step you need to select Transactional replication.

This is because P2P replication is a special case of Transactional replication and we will configure this later on. Select the tables you wish to replicate to all nodes in the next step and setup the correct SQL Server or Windows Domain account that has the appropriate permissions to talk to the database. Give the publication a name like MyP2PReplication and save it.

To configure this replication to be a P2P one, open the Properties dialog for the replication and go to the Subscription Options screen. Select True for the Allow P2P Subscriptions value and save the settings. Now, when you right-click the replication name you'll see a new option-Configure P2P Topology. Select this to bring up P2P configuration wizard.

Continue in the wizard till you reach a topology configuration screen. You can start adding nodes in this screen for P2P. Right click an empty part of the screen and select Add a new Peer Node. Select a database server (with right permissions) and then select the database name to use for replication. Make sure that you increment the Peer Originator ID for each node.

Once all the nodes are in place, right click any node and select Add a new Peer Connection. This brings up a rubber-banding arrow that moves with your mouse and attaches to other nodes.

Click on a node to create a P2P replication between that and the origin. Perform this for each node so that they talk to each other. You can also speed this up by selecting Connect to all displayed nodes which creates the topology for a server to all others. To create a mesh, perform this step with all nodes.

Continue on and enter the appropriate security information for access to each node's database. Once done, complete the wizard and wait till the topology is actually setup. Now right click Replication in any server and bring up Replication Monitor. This will show you whether the replication is correctly setup or not. To see the effect of replication, make some changes (insert, updates, deletes) in any table that you selected for replication in any server. Very soon you'll see the same change occuring in all other nodes as well.

Some very important things to keep in mind while setting up P2P replication.

1. P2P does not use Snapshots to create the 'base' image for databases across nodes. It is up to you to ensure that the databases start off with identical schemas and data before you setup the replication.

2. Primary conflicts can occur here. You need to ensure that if a table has data getting inserted at any node, primary keys are unique across nodes. This can be done by:

a. Using GUIDs as PKs

b. Using composite keys with one part of the key being unique to the server

c. Using Identity columns with start and step values to ensure no conflicts. For instance, one on server, the identity field can be 1+2 (so that it uses 1, 3, 5...) and on another it uses 2+2 (so that it uses 2, 4, 6 ...).

3. P2P is available only in the Enterprise Edition of SQL Server 2008.

Once you setup P2P it becomes very easy to manage data changes across multiple locations. This is a great new feature of SQL Server 2008 and can make the life of a database admin much better.