PostgreSQL continues to solidify its place as a top-tier database selection amongst workloads working on Kubernetes.
Within the ever-evolving world of cloud-native applied sciences, PostgreSQL continues to solidify its place as a top-tier database selection amongst workloads working on Kubernetes. In response to the Kubernetes within the Wild 2025 report, PostgreSQL now powers 36% of all database workloads working on Kubernetes—up 6 factors since 2022—signaling its rising recognition and rising belief among the many Kubernetes neighborhood1. Nevertheless, working data-intensive PostgreSQL workloads on Kubernetes has its personal set of challenges. These embrace managing Kubernetes primitives like StatefulSets and deployments, in addition to reaching optimum efficiency by configuring storage, replication, and database settings, however that is quick evolving to a simplified expertise.
We now present two choices for deploying stateful PostgreSQL workloads primarily based on efficiency wants. To help databases with stringent latency and scalable transaction necessities, you possibly can leverage Azure Container Storage to orchestrate Kubernetes quantity deployment on native NVMe to scale up IOPS whereas sustaining extraordinarily low sub-ms latency. For situations the place optimized price-performance is a precedence, Premium SSD v2 is the go-to selection. Moreover, working with CloudNativePG, we built-in a sturdy open-source operator for PostgreSQL to help a excessive availability database deployment mannequin on Azure Kubernetes Service (AKS). Our superior storage choices mixed with CloudNativePG make AKS a sturdy platform for high-performance PostgreSQL workloads.
Breakthrough PostgreSQL efficiency with native NVMe
For performance-critical PostgreSQL workloads, resembling these dealing with large concurrent transactions or demanding, low-latency information entry, native NVMe immediately connected to Azure Digital Machine (VM) SKUs is your finest guess. Utilizing native NVMe drives with Kubernetes was once sophisticated—it usually required organising RAID throughout the drives and manually managing static quantity orchestrators. Azure Container Storage successfully addresses this problem.
Azure Container Storage is a totally managed, container-native storage answer, designed particularly for Kubernetes. Builders can merely request a Kubernetes quantity, and Azure will dynamically provision storage backed by the out there native NVMe drives on AKS nodes. This offers PostgreSQL customers direct connect block storage IOPS and latency inside a managed, orchestrated cloud surroundings. Whether or not you’re powering fee methods, gaming backends, or real-time personalization engines, you get the most effective of each pace and ease. Azure Container Storage additionally helps Azure Disk and Elastic SAN (Preview), so you possibly can select backing storage with completely different sturdiness, scale, or price as your wants evolve—all beneath a constant, Kubernetes-native management aircraft.
Our benchmark outcomes have proven PostgreSQL reaching shut to fifteen,000 transactions per second (TPS) with single-digit millisecond end-to-end question latency with the Standard_L16s_v3 VM. When scaling as much as bigger VM SKUs like Standard_L64s_v3, we noticed TPS reaching as much as 26,000 whereas sustaining low latency. For extra particulars of our benchmark runs, confer with the comparability of storage choices part under.
Optimize price-performance with Premium SSD v2
Azure Premium SSD v2 affords an optimum stability of price-performance and a versatile deployment mannequin, making it particularly well-suited for manufacturing environments that must scale over time. With Premium SSD v2, you possibly can configure IOPS, throughput, and dimension independently—enabling PostgreSQL deployments to scale dynamically with demand whereas minimizing upfront prices and avoiding useful resource overprovisioning.
Whether or not you’re working multi-tenant SaaS platforms, manufacturing methods that scale with enterprise wants, or functions with spiky visitors, this flexibility results in actual financial savings with out sacrificing efficiency. With as much as 80,000 IOPS and 1,200 MB/s per quantity, Premium SSD v2 helps extremely demanding PostgreSQL workloads on an infrastructure that adapts to your app.
Comparability of storage choices
That can assist you assess the 2 storage choices outlined above, we performed benchmark runs utilizing the CloudNativePG operator setups on AKS with comparable core and reminiscence consumption, with each backing storage choices as the one variable: one leveraging native NVMe with Azure Container Storage, and the opposite utilizing Premium SSD v2 with Disk CSI driver.
For the primary configuration, we used Standard_D16d_v5 SKU and provisioned two Premium SSD v2 32 GiB disks every having 3000 IOPS and 125 MB/s throughput for log and information information. Within the second setup, we ran on Standard_L16s_v3 nodes with native NVMe storage included. The take a look at surroundings was configured to intently simulate a real-world manufacturing database state of affairs. TPS measures what number of particular person transactions (resembling INSERT, UPDATE, DELETE, or SELECT) a system can deal with per second. Latency refers back to the time delay between issuing a request to the database and receiving a response, which is very important for functions requiring real-time or near-real-time responsiveness, resembling monetary methods, on-line gaming, or high-performance analytics.
Native NVMe on Standard_L16s_v3 delivered 14,812 TPS with an common latency of 4.321 milliseconds. PremiumV2_LRS on Standard_D16ds_v5 recorded 8,600 TPS at 7.417 milliseconds latency. See pricing comparability under:
*Month-to-month prices are primarily based on the bottom 3000 IOPS and 125 MB/s throughput. You’ll be able to modify the efficiency (capability, throughput, and IOPS) of Premium SSD v2 disks at any time, permitting workloads to be price environment friendly whereas assembly workload dimension and efficiency necessities.
**With 3 VMs of L16s_v3, you get 11.52 TB of storage allotted by default that’s used to serve the volumes created for PostgreSQL workload. For different VM sizes within the L-Sequence household, the worth monthly and allotted storage will fluctuate.
For PostgreSQL workloads, the selection between utilizing native NVMe and Premium SSD v2 will depend on balancing efficiency, price, and information sturdiness. Native NVMe through Azure Container Storage affords extraordinarily low latency and excessive throughput, making it appropriate for performance-sensitive PostgreSQL deployments. The prices are increased with native NVMe, and there’s much less flexibility to scale independently of workload traits. Conversely, Premium SSD v2 offers higher price-performance effectivity and versatile scalability, making it a viable possibility for PostgreSQL situations that require dealing with elevated scale or functions having unpredictable surges in demand or utilization. By way of information sturdiness, Premium SSD v2 affords domestically redundancy by default, whereas for native NVMe, it is suggested to make use of a replica-based structure managed by CloudNativePG operator and an object storage-based backup strategy to stop information loss.
Constructed for prime availability with CloudNativePG on Azure Kubernetes Service
For groups deploying PostgreSQL in manufacturing, excessive availability and backups are non-negotiable. With the open-source CloudNativePG operator, a extremely out there PostgreSQL cluster on AKS can simply be deployed with:
- Constructed-in replication and automatic failover.
- Utility constant backup with native integration with Azure Blob Storage.
- Seamless integration with Azure Container Storage.
- Versatile storage choices: select Premium SSD v2 or native NVMe primarily based on workload wants.
Whether or not you’re supporting inside enterprise apps or customer-facing platforms, this provides you peace of thoughts with out the effort of hand-building customized excessive availability logic and separate backup workflows. Get began with deploying extremely out there PostgreSQL on AKS with CloudNativePG operator utilizing our step-by-step reference information.
Prepared for the long run
PostgreSQL is only one of many stateful workloads that organizations at the moment are confidently working on Azure Kubernetes Service. From databases to message queues, AI inferencing, and enterprise functions, AKS is evolving to fulfill the wants of persistent, data-heavy functions in manufacturing.
Whether or not you’re deploying Redis, MongoDB, Kafka, and even ML-serving pipelines with GPU-backed nodes, AKS offers the inspiration to handle these workloads with efficiency, consistency, and operational ease, together with clear end-to-end steerage.
With improvements like Azure Container Storage for native NVMe and Premium SSD v2 for scalable persistent storage, we’re making it simpler than ever to construct stateful functions which might be: dependable, performant, and price environment friendly for mission important workloads.
Modernize your information layer on Kubernetes right this moment. Whether or not you’re working PostgreSQL or any stateful tier, Azure delivers the efficiency and manageability to make it occur. Discover confirmed patterns and deployment choices within the AKS Stateful Workloads Overview.
