What is Azure Database for PostgreSQL?
We look at Microsoft's cloud implementation of the open source database
Azure Database for PostgreSQL is becoming increasingly popular since the launch of its Single Server or Hyperscale (Citus) in late 2019. However, in order to understand the fully managed database as a service, it's important to go back to its origins.
So, what do Microsoft, Apple, Instagram, Accenture, Reddit, Skype, Twitch, Spotify, and the International Space Station all have in common? Although they are all very different companies working across a range of sectors, they all run their essential workloads using the same open source database system – PostgreSQL.
Considered to be the most sophisticated open source database system, PostgreSQL was launched almost exactly 25 years ago as a successor to the Ingres database developed in the 1970s by scientists from the University of California, Berkeley. So what does this one open source database have in common with Microsoft’s cloud computing service?
While the original PostgreSQL is a free and open-source relational database management system, Azure Database for PostgreSQL is the tool which allows users to build massively scalable PostgreSQL applications. Based in Microsoft’s cloud infrastructure, Azure Database for PostgreSQL makes it possible to develop applications using PostgreSQL technology using all of its elements, from tools and drivers to libraries, without actually needing to take control of the instances. The service, available in different deployment models, can be run on premises on any infrastructure the customer chooses, with the benefit of elastic scale, unified management, and a cloud billing model.
In this article, we will break down the tiers and pricing models of Azure Database for PostgreSQL, as well as help you decide which deployment model is best for your organisation.
Azure Database for PostgreSQL – Single Server
The main deployment method for PostgreSQL databases on Azure is the Single Server model, and very closely resembles the conventional PostgreSQL. Users can also take advantage of the on-premise version to create a single PostgreSQL server as the main host for multiple databases.
The Single Server model also offers a fully managed service that can boast 99.99% performance guarantee, and includes several additional features, such as encryption.
This configuration can also be handled through the Azure portal as well as the Azure command line interface (CLI), as can the majority of Microsoft’s cloud services. The Single Server deployment can also let developers how a single database per server, which means they can make the most of resource usage and even share resources across several databases.
Administrators on Azure Database for PostgreSQL don’t always have full superuser permissions, unfortunately, with the highest privileges available going to the azure_pg_admin. The managed service will always retain the superuser attributes, assigned to azure_superuser, and regular users aren’t able to access this role or its privileges.
Azure Database for PostgreSQL - Hyperscale (Citus) (preview)
If you have a database which is over 100GB and requires maximum performance, you might want to consider Azure's newest PostgreSQL deployment model. Also known as 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.
Understanding your open source risk
Open source libraries can introduce vulnerabilities to your code
Download nowPricing
There are three pricing tiers available for the single server version of Azure Database for PostgreSQL, with each setting out the various provisions that a customer may expect from these packages. These are termed Basic, General Purpose, and Memory Optimised.
While all tiers offer computing power provisioned in virtual cores, or vCores, and each uses fifth-generation chips based on Intel’s Broadwell 2.3 GHz processors, the number of cores and the memory available varies.
The entry-level tier, for instance, offers either one or two rentable vCores, combined with 2GB memory, for £0.026 per hour and £0.051 per hour each. General Purpose and Memory Optimised configuration both offer far more memory and vCores, and are priced more steeply, with the maximum configuration available for £4.179 per hour.
Elements of the servers’ configuration can be changed and adjusted after they’re established, however, and there are a number of factors which combine to influence how much the system will end up costing. You could, for example, ramp up the number of vCores deployed, the storage capacity, and the length of time that backups are retained. Users can also interchange between the different pricing tiers, between General Purpose and Memory Optimised setups, for example.
The final pricing also depends on the type of package you take out and how you want to pay for it, whether it be pay-as-you-go, or annual billing, which generally cost less as you pay more upfront. This annual discounting is not available on the basic level, however. The full breakdown of available tiers is listed below, with fuller pricing information available on through Microsoft.
Understanding the tiers
| Pricing tier | Basic | General Purpose | Memory Optimised |
| Compute generation | Gen 5 | Gen 5 | Gen 5 |
| vCores | 1, 2 | 2, 4, 8, 16, 32, 64 | 2, 4, 8, 16, 32 |
| Memory per vCore | 2GB | 10GB | 20GB |
| Storage capacity | Up to 1TB | Up to 16TB | Up to 16TB |
| Storage type | Azure Standard Storage | Azure Premium Storage | Azure Premium Storage |
Each tier provides a free back up of up to 100% of your provisioned server storage. You can increase your backup retention period, however, this will increase the amount of backup storage consumed by the server, and so you will be charged extra.
Basic
The entry-level tier is primarily designed for low-priority workloads that don't require a great deal of performance. This can include test/dev environments or intermittently-accessed applications.
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.
Understanding your open source risk
Open source libraries can introduce vulnerabilities to your code
Download nowBenefits 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 virtual machines 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.
