May 25, 2024


This publish was co-authored by Maneesh Sah, Company Vice President, Azure Storage Engineering.

Container is the brand new digital machine (VM). Whether or not you’re a CTO, enterprise architect, main DevOps groups, or constructing purposes, you might have already launched into the journey to containerize your purposes or are raring to go—to maximise the advantages of scale, flexibility, and value. With Kubernetes on the helm, containers have quickly change into a hotbed of innovation and a essential space of transformation for enterprises and startups alike. After the preliminary deal with stateless containers, working excessive scale stateful workloads on containers has now change into the norm. To run business-critical, enterprise-grade purposes on Kubernetes within the cloud, clients want extremely scalable, cost-efficient, and performant storage—built-for and with intrinsic assist for containers. At present, we’re excited to announce the preview of Azure Container Storage, the trade’s first platform-managed container native storage service within the public cloud, offering finish to finish storage administration and orchestration for stateful purposes to run effectively at scale on Azure.

Why Azure Container Storage?

With speedy adoption of Kubernetes, we see a surge of manufacturing workloads, each cloud-first in addition to app modernization, that want container-native persistent storage for databases (resembling MySQL), large knowledge (resembling ElasticSearch), messaging purposes (resembling Kafka), and steady integration and steady supply (CI/CD) methods (resembling Jenkins). To run these stateful purposes, clients want operational simplicity to deploy and scale storage tightly coupled with the containerized purposes. Clients as we speak, nevertheless, want to decide on between utilizing VM centric cloud storage choices, retrofitted to containers, or deploying and self-managing open-source container storage options within the cloud—main to very large operational overhead, scaling bottlenecks, and excessive price.

To offer clients with a seamless end-to-end expertise, container native storage must allow:

  • Seamless quantity mobility throughout the cluster to maximise pod availability with out bottlenecks on quantity attaches and deletes.
  • Speedy scaling of huge variety of volumes as utility pods scale up or scale out as wanted.
  • Optimum price-performance for any quantity sizes, particularly small volumes that require increased enter/output operations per second (IOPS).
  • Easy and constant quantity administration expertise throughout backing storage varieties to match workload necessities, resembling extraordinarily low latency ephemeral disks versus persistent or scalable distant storage.

Azure Container Storage addresses these necessities by enabling clients to focus their consideration on working workloads and purposes somewhat than managing storage. Azure Container Storage is our first step in direction of offering a transformative storage expertise. As a essential addition to Azure’s suite of container companies, it is going to assist organizations of all sizes to streamline their containerization efforts and enhance their total storage administration capabilities.

Leveraging Azure Container Storage

Azure Container Storage introduces the concept of Container Storage Pools, where a pool is a grouping of storage resources that are presented as a storage entity for your AKS cluster. Existing Azure Storage offerings like ephemeral disks, Azure Disks, and Managed storage (backed by Azure Elastic SAN) can all serve as resources backing a storage pool. You can create a persistent volume (PV) in this storage pool and mount it to your pods, to store data.

Azure Container Storage is a purpose-built, software-defined storage answer that delivers a constant management airplane throughout a number of backing storage choices to fulfill the wants of stateful container purposes. This totally managed service offers a quantity administration layer for stateful container purposes enabling storage orchestration, knowledge administration, Kubernetes-aware knowledge safety, and rule-based efficiency scaling.

Aligning with open-source container native storage approaches, Azure Container Storage runs microservices-based storage controllers in Kubernetes, to summary the storage administration layer from pods and backing storage, enabling portability throughout Kubernetes nodes and talent to mount totally different storage choices as proven within the determine.

Azure Container Storage elements embrace:

  • A Storage Pool, which is a set of storage sources grouped and introduced as a unified storage entity on your AKS cluster.
  • An information companies layer, answerable for replication, encryption, and different add-on performance absent within the underlying storage supplier.
  • A protocol layer, which exposes provisioned volumes through NVMe-oF protocol to utility pods.

With this method Azure Container Storage affords a number of differentiated experiences to clients on Azure, together with:

Decreasing the entire price of possession (TCO) by offering means to scale IOPS on smaller quantity sizes, to assist containerized purposes which have dynamic and fluctuating enter/output (IO) necessities. That is enabled utilizing shared provisioning of capability and efficiency on a storage pool, which may be leveraged by a number of volumes. With shared provisioning, clients can now maximize efficiency throughout utility containers whereas conserving TCO down. As a substitute of allocating capability and IOPS per persistent quantity (PV), which generally results in overprovisioning, clients can now create PVs and dynamically share sources from a Storage Pool.

Speedy scale-out of stateful pods, achieved utilizing distant community protocols like NVME-oF and iSCSI to mount PV, enabling easy scaling on AKS throughout Compute and Storage. That is particularly helpful for container deployments that begin small and iteratively add sources. Responsiveness is vital to make sure that purposes should not starved or disrupted, both throughout initialization or scaling in manufacturing. Moreover, utility resiliency is vital with pod respawns throughout the cluster requiring speedy PV motion. Leveraging distant community protocols permits us to tightly couple with the pod lifecycle to assist extremely resilient excessive scale stateful purposes on AKS.

Simplified constant quantity administration interface backed by native and distant storage choices enabling clients to allocate and use storage through the Kubernetes management airplane. Which means clients can leverage ephemeral disks, Azure Disks in addition to Azure Elastic SAN through a unified administration interface to fulfill workload wants. As an example, ephemeral storage could also be preferable for Cassandra to realize the bottom latency, whereas Azure Disks is appropriate for PostgreSQL or different database options. This unified expertise offered by Azure Container Storage simplifies the administration of persistent volumes, whereas delivering a complete answer to handle the broad vary of efficiency necessities of assorted containerized workloads.

Absolutely built-in day-2 experiences, together with knowledge safety, cross-cluster restoration, and observability offering operational simplicity for purchasers who must create buyer scripts or sew collectively disparate instruments as we speak. Clients can orchestrate Kubernetes-aware backup of the persistent volumes built-in with AKS usually out there to streamline the end-to-end experiences for working stateful container workloads on Azure.

Seek advice from the technical neighborhood weblog publish for extra particulars about Azure Container Storage.

Getting began

Signal as much as take part within the preview and deploy your first stateful container utility as we speak.

Seek advice from the Azure Container Storage documentation to be taught extra in regards to the service.

We’re assured that this new provide will considerably speed up app modernization and cloud migration. We look ahead to listening to your suggestions. Please e-mail us at with any questions.


Source link