The evolution of Kubernetes as a portable, extensible, and distributed computing platform makes it a popular choice for modern cloud-native applications designed to run on a commodity software-defined infrastructure. Organizations are increasingly leveraging Kubernetes for stateful applications, and they require a storage solution that can manage state in a reliable way with the highest levels of availability and scale. The solution also needs to be simple enough for Kubernetes admins and application owners to operate on their own.
Speaking of simplicity, many cloud-native applications need files to store their data. A storage solution that natively supports files is easier to manage because it mitigates the challenges of configuring and running isolated local file systems. In addition, cloud-native applications are designed to scale horizontally with the infrastructure. Therefore, a storage solution that can scale infinitely with the underlying infrastructure is like insurance for application owners – they can start small and scale on demand.
Introducing Astra Data Store
Today we are excited to announce a preview of NetApp® Astra™ Data Store, the latest addition to NetApp’s Astra product portfolio. Astra Data Store decodes the core challenges of running stateful applications on Kubernetes.
Kubernetes-native shared file services with integrated data management
Astra Data Store is one of the first in the industry to provide Kubernetes-native shared file services integrated with data management functions offered as a fully software-defined solution. Astra Data Store gives today’s cloud-native applications a shared filesystem with ReadWriteMany/ReadOnlyMany access so that they can scale horizontally by consuming the same persistent volume from multiple nodes. This also provides the flexibility to scale the application performance and capacity needs independently. Legacy applications (that were written based on shared filesystem architectures) can be migrated more easily to Kubernetes while still maintaining the same levels of data availability. Further, the integrated data management functions not only present the simplicity of managing the application data lifecycle from the Kubernetes control plane but also leverage the block level storage efficiencies of the filesystem data layout.
Astra Data Store truly embraces the challenges of running in a commodity-software-defined infrastructure and provides topology aware data protection to protect data against component failures (node, drive, etc.) in the cluster ecosystem. Application data is protected not only across a node failure but also across the failure of an entire rack when the solution is configured across racks. Astra Data Store can operate with data-center-grade flash drives and can rebuild the data locally when a drive fails within a node. This process reduces the blast radius of the repair operation and maintains the application quality of service. Astra Data Store software can also detect block-level inconsistencies and repair both file system data and metadata – automatically.
Scale-out architecture with linear scaling
Finally, Astra Data Store is built on an elastic and scalable architecture that can scale according to the application’s needs. It is a highly scalable distributed system that enables linear scaling of both capacity and storage performance. The distributed algorithms manage load uniformly across the cluster, preventing any hot spots. The on-disk data structures not only enable media-friendly write transactions but also apply a combination of probabilistic and deterministic search methods for fast lookups during read operations and garbage collection.
The key architecture tenets of Astra Data Store provide the reliability, availability, and scale required for stateful Kubernetes apps, as well as the Kubernetes-native simplicity required by Kubernetes practitioners, application owners, and developers. These capabilities are essential for increasing the adoption of cloud-native technologies.
You can find details on the Astra Data Store announcement here.