Replication Availability Enhancements
Replication has made availability enhancements in the following areas:
Replication of schema changes
Resumable snapshot delivery
Peer-to-peer transactional replication
Improved support for Replication Agents
Logical record replication
Replication of Schema Changes
In SQL Server 2000, columns could be added to and dropped from published tables using the procedures sp_repladdcolumn (Transact-SQL) and sp_repldropcolumn (Transact-SQL). In SQL Server 2005 a much broader range of schema changes can be replicated without using special stored procedures. DDL statements are issued at the Publisher and are automatically propagated to all Subscribers. For more information, see Making Schema Changes on Publication Databases.
Resumable Snapshot Delivery
Improvements have been made to snapshot generation and application, including the automatic resumption of snapshots interrupted during delivery. If the snapshot transfer is interrupted at any point, it automatically resumes and does not resend any files that have already been completely transferred. No special options are required to take advantage of this functionality.
Peer-to-Peer Transactional Replication
In SQL Server 2000, transactional replication supported hierarchical topologies in which a Publisher owned the data that was replicated to Subscribers. Transactional replication with updating subscriptions supported Subscriber-side updates, but the Subscribers were classified as different types of participants in replication than the Publishers. In SQL Server 2005, a new peer-to-peer model is introduced that allows replication between identical participants in the topology. The new support is targeted at customers running server-to-server configurations that also require moving the roles between replicated nodes dynamically for maintenance or failure management. For more information, see Peer-to-Peer Transactional Replication.
Improved Support for Replication Agents
Reliability and error recovery have been improved for replication agents. Contention between agents and jobs is decreased, and agents automatically retry on network errors, deadlock conditions, and query timeouts. For more information about agents, see Replication Agents Overview.
Logical Record Replication
By default, merge replication processes changes on a row-by-row basis. The logical records feature allows merge replication to treat as a single unit a set of related rows, such as a parent row in a SalesOrderHeader table and its child rows in SalesOrderDetails. Using this feature ensures that, regardless of network reliability or any other factors, related sets of records will always be processed in their entirety at the same time at a Subscriber. For more information, see Grouping Changes to Related Rows with Logical Records.