Lead Image © Golkin Oleg, 123RF.com

PostgreSQL Replication Update

Data Duplication

Article from ADMIN 40/2017

By Hans-Jürgen Schönig

High availability, replication, and scaling are everyday necessities in the database world. What features does PostgreSQL offer in this context, and how good are they?

If you have anything to do professionally with the topic of replication as a PostgreSQL consultant [1], you might easily gain the impression that colleagues and clients see only distributed, synchronous, multimaster replication as true replication. But are they right?

Synchronous vs. Asynchronous

Generally speaking, replication can be classified in different ways (e.g., synchronous or asynchronous) (Figure 1). In the case of synchronous replication, a transaction is not completed until it is acknowledged by the service. The advantage, particularly in the case of a crash, is that the data must have arrived on at least two systems. The drawback is that the data must be written to at least two systems. The advantage is therefore a disadvantage because the double write operations cost time.

Figure 1: In the case of asynchronous replication, a time offset occurs when executing the transactions.

Single- and Multimaster

The distinction between single-master and multimaster is important. Many users choose multimaster because they can use it to distribute the read load, although it does not make writing faster because both

...

Use Express-Checkout link below to read the full article (PDF).