Accelerating AI and databases with Azure Container Storage, now 7 times faster and open source

The adjacent large merchandise of Azure Container Storage delivers up to 7 times higher IOPS and 4 times little latency compared to erstwhile versions.

More companies than ever earlier are choosing to tally stateful workloads—such arsenic relational databases, AI inferencing, and messaging queues—on Kubernetes. For developers gathering connected Kubernetes, retention show has ne'er been much important.

Today, we’re announcing the adjacent large merchandise of Azure Container Storage – v2.0.0. Compared to anterior versions, it delivers up to 7 times higher IOPS, 4 times little latency, and improved assets efficiency. With built-in enactment for section NVMe drives, Azure Container Storage present delivers our fastest, astir almighty Kubernetes retention level connected Azure. It’s present besides wholly escaped to use, and disposable arsenic an open-source version for installation connected non-AKS clusters. Whether you’re moving stateful applications successful production, scaling AI workloads, oregon streamlining dev/test environments, this large release’s show volition springiness your workloads a sizeable boost.

What’s Azure Container Storage?

Before we dive into the latest enhancements, let’s instrumentality a infinitesimal to revisit what Azure Container Storage is and however developers tally stateful workloads connected Kubernetes with speed, simplicity, and reliability.

Azure Container Storage is simply a cloud-native measurement absorption and orchestration work specifically designed for Kubernetes. It integrates seamlessly with AKS (Azure Kubernetes Service) to alteration provisioning of persistent volumes for production-scale, stateful workloads.

Azure Container Storage’s imaginativeness is to service arsenic the unified artifact retention orchestrator for Kubernetes workloads connected Azure, providing a accordant acquisition crossed aggregate retention backends for simplified measurement orchestration via Kubernetes APIs. This v2.0.0 merchandise focuses specifically connected breakthrough show with section NVMe storage, bringing enterprise-grade show with cloud-native simplicity. Later this year, we’ll beryllium debuting enactment for Azure Container Storage to integrate with Elastic SAN.

Azure Container Storage delivers optimized show and ratio with low-latency retention for precocious throughput stateful applications, on with built-in orchestration and automation that allows Kubernetes to negociate retention pools, persistent measurement lifecycles, snapshots, and scaling—all without switching contexts oregon managing idiosyncratic CSI (container retention interface) drivers.

What’s new?

There’s rather a spot to unpack here, truthful let’s instrumentality a deeper dive into immoderate of the cardinal benefits that Azure Container Storage v2.0.0 delivers:

Pricing changes

As before, you’ll proceed to wage for the underlying retention backend you use. But Azure Container Storage versions 2.0.0 and beyond volition nary longer complaint a per-GB monthly interest for retention pools larger than 5 TiB for some our archetypal enactment managed and open-source version, making the work present wholly escaped to use. Provision arsenic overmuch retention arsenic you request without worrying astir further absorption fees. This means you get enterprise-grade retention orchestration and breakthrough show without immoderate further work costs—just axenic worth for your Kubernetes workloads.

Enhanced show with reduced assets consumption

This merchandise of Azure Container Storage is optimized specifically for section NVMe drives provided with a assortment of VM families. This absorption unlocks the fastest imaginable I/O show for your astir demanding workloads portion reducing infrastructure costs.

Perhaps astir exciting, this latest mentation of Azure Container Storage connected section NVMe is present faster than ever before. We’ve rebuilt our architecture from the crushed up—from the kernel level to the power plane—to propulsion the limits of our retention orchestrator. This melodramatic velocity betterment comes with an arsenic awesome simplification successful clump assets consumption. Previously, Azure Container Storage connected section NVMe had 3 show modes that could devour 12.5%, 25%, oregon 50% of your node pool’s CPU cores. Azure Container Storage v2.0.0 nary longer has show tiers. Instead, it delivers superior show portion utilizing less resources than adjacent our erstwhile lowest-impact setting. This translates straight to outgo savings—you get amended show portion freeing up CPU capableness for your applications to execute adjacent faster.

Let’s look astatine the benchmarks. On fio (Flexible I/O Tester), the open-source manufacture modular for retention testing, Azure Container Storage connected NVMe delivers astir 7 times higher IOPS and 4 times little latency compared to the erstwhile version.

But however does this construe to existent workloads? We tested our ain PostgreSQL for AKS deployment guide and recovered that PostgreSQL’s transactions per 2nd improved by 60% portion cutting latency by implicit 30%. For database-driven applications, this means faster query responses, higher throughput, and amended idiosyncratic experiences.

All successful all, Azure Container Storage delivers a important show boost for I/O-demanding workloads retired of the container without further configuration needed, offering developers a elemental yet almighty instrumentality successful their cloud-native arsenal.

Accelerated AI exemplary loading and KAITO Integration

For AI and instrumentality learning workloads, exemplary loading clip tin beryllium a important bottleneck. Azure VMs equipped with GPUs person section NVMe drives available. With the latest NVMe enhancements successful the caller v2.0.0 version, Azure Container Storage takes vantage of this hardware by dramatically accelerating exemplary record loading for AI inferencing workloads. With our recent integration with KAITO, the archetypal Kubernetes-native controller for automating AI exemplary deployment, you tin present deploy and standard AI models faster than ever, reducing time-to-inference and improving wide AI exertion responsiveness.

Above: Azure Container Storage providing accelerated NVMe-backed retention for exemplary files

We loaded Llama-3.1-8B-Instruct LLM and recovered a 5 times improvement in exemplary record loading velocity with Azure Container Storage v2.0.0, compared to utilizing an ephemeral OS disk.

More flexible scaling options

Azure Container Storage antecedently required a minimum of 3 nodes erstwhile utilizing ephemeral drives. It present works with clusters of immoderate size, including single-node deployments. This flexibility is peculiarly invaluable for applications with robust built-in replication oregon backup capabilities, improvement environments, and borderline deployments wherever you request high-performance retention without the overhead of larger clusters. The elimination of minimum node requirements besides reduces costs for smaller deployments portion maintaining the aforesaid high-performance capabilities.

Open root and assemblage support

We admit however important the open-source assemblage is to the wellness and tone of the Kubernetes ecosystem. Azure Container Storage mentation 2.0.0 is present built connected our recently created open-source repositories, making it accessible to the broader Kubernetes community.

Whether you request the Azure-managed mentation for seamless AKS integration oregon similar the assemblage open-source mentation for self-hosted Kubernetes clusters, you get the aforesaid large merchandise and features. The open-source attack besides means easier installation, greater transparency, and the quality to lend to the project’s evolution.

Explore our open-source repository (local-csi-driver), and larn much astir our related artifact retention products:

• Azure Container Storage enabled by Azure Arc
• Use Container Storage Interface (CSI) operator for Azure Disk connected Azure Kubernetes Service (AKS)

In summary

This large update to Azure Container Storage delivers a faster and leaner high-performance Kubernetes retention platform. Here’s what you get:

• Included retired of the box: This merchandise focuses connected ephemeral drives (local NVMe and impermanent SSD) provided with select VM families, including storage-optimized L-series, GPU-enabled ND-series, and general-purpose Da-series.
• Enhanced workload support: Optimized for demanding applications similar PostgreSQL databases and KAITO-managed AI exemplary serving.
• Superior performance: 7 times betterment successful read/write IOPS and 4 times simplification successful latency, with 60% amended PostgreSQL transaction throughput.
• Open source: Built connected open-source foundations with community repositories for easier installation connected immoderate Kubernetes cluster.
• Flexible scaling: Deploy connected clusters with arsenic fewer arsenic 1 node—no minimum clump size requirements.
• Zero work fees: Completely escaped to usage for each retention excavation sizes—you lone wage for underlying storage.

Getting started

Ready to acquisition the show boost? Here are your adjacent steps:

• New to Azure Container Storage? Start with our comprehensive documentation.
• Deploying circumstantial workloads? Check retired our updated deployment usher for PostgreSQL.
• Want the open-source version? Visit our GitHub repository for installation instructions.
• Have questions oregon feedback? Reach retired to our squad astatine AskContainerStorage@microsoft.com.

Regardless of your workload, Azure Container Storage provides the show and easiness you expect from modern cloud-native storage. We’re excited to spot what you build—and we’d emotion to perceive your feedback. Happy hacking!