Chapter 2
Introduction

If you are new to Empress Replication, after reading this chapter, you may try the [Quick Start Tutorial] just to perform a basic replication.

2.2. What is Empress Replication

Figure 2-1: An example of a Replication World

A Master Table is an update-able database table that serves as a source of data in a Replication World. A Replicate Table is a read-only database table that gets a copy of data that is updated in a Master Table.

In a Replication World, all data updates are made to a Master Table and are later propagated to Replicate Tables. In [Figure 2.1] data from a Host A have to be propagated to Hosts B, C and D. A propagation of data is done by means of a synchronization. In the process of copying data during synchronization, the origin of data is called Replication Master Table (or RMT) and the copy is called Replication Replicate Table (or RRT).

Note that a table in a Replication World, in its nature, can be a Master or a Replicate Table, but can play two different roles during a synchronization process: it can act as a Replication Master Table (RMT) or as a Replication Replicate Table (RRT). For example, the Replicate Table in the host C in [Figure 2.1] acts as the RRT during the synchronization requested from the Host C, but acts as the RMT for the synchronization requested from the Host D.

Replication Synchronization is a process of updating an RRT with a changed data of its RMT since the last successful synchronization. Synchronization request is made by a Synchronization Client (Empress Replication Synchronization Utility - emprepsync). Synchronization Client attempts to connect to an Empress Replication Master Server, which provides replication (i.e. synchronization) services to its clients. Server listens to synchronization requests from its clients and provides a copy of data from an RMT. In [Figure 2.1] two Replication Master Servers propagate data to all the Hosts: the Server in the Host A and the Server in the Host C.

A synchronization request can be made manually, or can be automated. For this operation, an RRT side does not have to be always connected to a host machine running Replication Master Server, but only for a duration of synchronization. Empress replication is an asynchronous replication. Hence, there will always be a lag between RMT and its RRT defined by a frequency of synchronization.

When created, a Replicate Table inherits a table structure of its RMT. In general, all Replicate Tables in a Replication World have the same structure as a Master Table. Furthermore, since a Master Table is the only update-able table in a Replication World, all data comes from the same source. For the sake of performance, creating a Replicate Table provides an option for the initial loading of the RMT data, before the indices are created.

Since Empress Replication propagates data changes, it is Data-based Replication (instead of being Log-based Replication, when the commands are re-executed on the replicates).

Subset Replication is a special feature of Empress Replication. A replicate table is normally created as a replica of an entire RMT, but it may also be created as a subset of an RMT by specifying a conditional expression (using an SQL WHERE Clause). Subset Replication duplicates rows from an RMT that satisfy the predetermined condition.

2.3. Why to use Empress Replication

Higher data availability:

An application can access a local source of data instead of a remote one. An application can access an alternative data source instead of only one, which may be occasionally inaccessible.

Reducing network traffic by providing a local copy of data:

Access to a local copy of data, reduces network traffic.

Off loading a table/record lock contention between database applications:

An application can access an alternative data source instead of only one. This decreases the likelihood of many applications accessing same data.

Distributing system load to multiple locations:

Queries can be performed on alternative locations containing a copy of data, so that a query load on a single location is reduced.

A recovery from system failures:

In the case of system failure, there exists another option for a table recovery, since a backup of original data lives in several sites.

Providing data copies through a network:

Data can be automatically distributed to remote sites in a network, instead of using manual methods or transferring physical media storage such as magnetic tape.