Cari Blog Ini

02 Juli 2012

Implementing Transactional Replication

Transactional replication can be implemented in a one-to-many scheme or in a many-to-one scheme.
Often, transactional replication is implemented over a wide area network (WAN).


* One-to-Many Replication
Most implementations of transactional replication involve a one-to-many replication scheme.
In this type of implementation, one table is published to one or more subscribers.

* Many-to-One Replication
In a many-to-one replication scheme,
one database subscribes to more than one subscription.
This is not the most common replication scheme, but it is nevertheless widely used.
Because transactional replication works by reading the transaction log on the publisher
and applying inserts, updates, and deletes to the subscriber,
it is perfectly suited for this replication scheme.
The only potential drawback is the fact that the subscribing table
must have a primary key defined on it.
As long as your data sources do not violate this primary key,
the many-to-one replication scheme will function properly.

* Replication over a WAN
Replication over a WAN is not only possible, but also quite common.
If you replicate data over a WAN,
you should continuously monitor the distribution history
and look for excessive replication times caused by network bandwidth limitations.
In addition, if possible,
configure replication such that only the Distribution Agent connects over the WAN.
That is, configure the system such that the distributor and publisher
are on the same side of the WAN.


Source:
Microsoft SQL Server 2000 Administrator's Companion Book