What is Azure Database for PostgreSQL?

PostgreSQL is the most advanced open source database system and it's used by a number of organisations to run mission-critical workloads - including Microsoft's Azure cloud service.

Azure Database for PostgreSQL is slightly different, however, as it's a managed implementation of the service running on Azure cloud infrastructure. The idea is to allow organisations to develop applications with PostgreSQL, also with the native tools, drivers and libraries that come with it, quickly and without taking control of the instances. 

Users can deploy it either as a Single Server or Hyperscale (Citus) which is currently in preview mode, pending full release. 

Azure Database for PostgreSQL - Single Server

The Single Server model is the primary deployment method for PostgreSQL databases on Azure. The configuration is almost identical to PostgreSQL; in its on-premise version, users can mockup a single PostgreSQL server as the main host for multiple databases. It's a fully managed service and offers a four-nines performance guarantee. It also includes built-in back and encryption features. 

Like the majority of Microsoft's cloud tools, the configuration can be done in the Azure portal or the Azure CLI and the single server option lets developers host a single database per server. This means they can maximise resource usage and or share those resources with multiple databases. 

Azure Database for PostgreSQL admins do not have full superuser permissions, however; the highest privileged user role available to users of the service is the azure_pg_admin role. Instead, superuser attributes are assigned to the azure_superuser, which belongs to the managed service. Service users do not have access to this role or its associated privileges

Azure Database for PostgreSQL - Hyperscale (Citus) (preview)

If you have a very large (100GB-plus) database that requires maximum performance, you may be interested in Azure's newest PostgreSQL deployment model. Officially dubbed the 'Hyperscale (Citus) (preview)' hosting type, this method uses technology from Citus Data, a company that Microsoft acquired in January 2019.

Hyperscale (Citus) uses database sharding technology, which splits data into smaller component parts and distributes them across a large number of compute nodes which are grouped together into a cluster. This cluster offers more storage capacity and CPU utilisation than a standard single-server PostgreSQL deployment would be able to.

Big companies like Facebook and Google use database sharding within their data centres, but one of the advantages of Hyperscale (Citus) is that sharding is handled automatically, without the tenant application needing to be taught how to do it. The system parallelises SQL queries and other operations across available servers, with a central 'coordinator node' handling query routing, aggregation and planning, and 'worker nodes' storing data.

When the coordinator receives a request from the application, it routes the query to the relevant worker node(s), depending on where the data in question is stored. Caveat emptor, however; as Hyperscale (Citus) is in public preview, it does not offer an SLA at the time of writing.

Pricing

The single server version of Azure Database for PostgreSQL comes with different pricing tiers, namely ‘Basic’, ‘General Purpose’ and ‘Memory Optimised’.

Each tier sets out the various provisions that a customer can expect. Each tier has compute provisioned in virtual cores, known as ‘vCores’ in the documentation, and each uses Gen 5 CPUs based on Intel’s Broadwell 2.3GHz processors. 

However, the number of vCores, and the memory provisioned, will vary depending on the tier. For example, with Basic, you can rent either 1 or 2 vCores, each provided with 2GB of memory, for £0.026 per hour and £0.051 per hour respectively - although this increases all the way up to 64 vCores, with 320GB of memory, for £4.179 per hour.

You can tweak certain elements of your servers' configuration after they're created, such as increasing optimised memory - but there are a number of additional factors which will influence your monthly bill. This includes the amount of compute capacity, the number of vCores, amount of storage and backup retention period, as well as switching between the general purpose and memory optimised pricing tiers.

Whether you choose to pay as you go or opt for yearly payment will also affect the final price, although yearly discounts are not available as part of the Basic tier.

You can see a full breakdown of the prices below. The number of databases per server has no direct effect on prices.

Understanding the tiers

General Purpose

As the name suggests, the general purpose tier is where most workloads will naturally fall, and encompasses most enterprise PostgreSQL use-cases. It offers a balance of performance and economy.

Memory Optimised

The highest tier is reserved for applications like financial transaction databases or analytics engines where low latency is paramount. For this reason, it makes heavy use of in-memory computing.

While the Basic tier does not provide an IOPS guarantee, other tiers have IOPS scale with the provisioned storage size in a 3:1 ratio.

Benefits of Azure Database for PostgreSQL

The database cloud service has a number of advantages.

Built-in high availability: The service provides built-in high availability with no additional setup, configuration or extra cost. This means there is no need to set up further VMs and configure replication to guarantee high availability for a PostgreSQL database.

Security: All data including backups are encrypted on disk by default. Also, the service has SSL enabled by default, so all data in-transit is encrypted.

Scalability: The service allows users to scale compute on the fly without application downtime in one step.

Automated backups: Users do not need to independently manage storage for backups. The service offers up to 35 days retention for automated backup.