An introduction to PGD Essential v6.0.1

EDB Postgres Distributed (PGD) Essential is a simplified version of PGD Expanded, designed to help you get started with distributed databases quickly and easily. It provides the core features of PGD, enabling high availability and disaster recovery without the complexity of advanced configurations.

At the core of PGD are data nodes, Postgres databases that are part of a PGD cluster. PGD enables these databases to replicate data efficiently between nodes, ensuring that your data is always available and up-to-date. PGD Essential simplifies this process by providing a standard architecture that is easy to set up and manage.

The standard architecture is built around a single data group, which is the basic architectural element for EDB Postgres Distributed systems. Within a group, nodes cooperate to select which nodes handle incoming write or read traffic, and identify when nodes are available or out of sync with the rest of the group. Groups are most commonly used on a single location where the nodes are in the same data center and where you have just the one group in the cluster, we also call it the one-location architecture.

Standard/One-location architecture

The one-location architecture consists of a single PGD cluster with three nodes. The nodes are located in the same data center or region. Ideally they are in different availability zones, but that isn't required. The nodes are connected to each other using a high-speed network.

The nodes are configured as a data group which means that they replicate data to each other within the same group. While PGD can handle multiple writers in a network, this requires more advanced conflict management and is not supported in PGD Essential.

Therefore, in the standard architecture, one node is designated as the write leader node, which handles all write transactions. The other nodes in the group are read-only nodes that replicate data from the write leader.

The write leader node is one node selected by the nodes in the group to handle all the writes. It is responsible for accepting write transactions and replicating them to the other nodes in the group. If the write leader node fails, the other nodes in the group will elect a new write leader node.

Applications can connect to any node in the cluster using PGD's Connection Manager ports which runs on every data node. It will automatically route read and write transactions to the write leader. It can also route read only transactions to the other nodes in the group.

In this diagram, you can see the applications connecting to the PGD cluster through the Connection Manager ports. The Connection Manager is responsible for routing the read and write transactions to the appropriate nodes in the group. The write leader is responsible for handling all write transactions and is shown in at the top in AZ1 in green.

The other nodes in the group are read-only nodes that replicate data from the write leader. Applications connecting to the read-only nodes Connection Manager read/write ports will have their queries and changes routed to the write leader. All the time, the nodes are talking to each other replicationing data to ensure they are in sync.