NetApp today launched Project Astra, a storage initiative for Kubernetes that promises to both optimize I/O performance and ensure data portability across clusters.
Eric Han, vice president of product management for NetApp’s Cloud Data Services business unit, says that as stateful applications become more widely deployed on Kubernetes clusters, there is now a need for data and storage management platforms that run natively on Kubernetes clusters that are designed to dynamically scale up and down as required.
NetApp, he says, is working with the Kubernetes community to create a storage platform that will automatically discover applications that can then share access to a common pool of data running natively on the Kubernetes cluster. That data can then be moved to any distribution of Kubernetes running on-premises or in the cloud.
That approach will advance the current level of access to persistent storage in Kubernetes environments, which Han notes today is largely achieved by accessing external storage systems via the Container Storage Interface (CSI) defined by the Kubernetes community.
Containers initially were employed to mainly build stateless applications. However, in the last year, the number of stateful container applications being deployed has sharply increased, most of which are now being deployed on Kubernetes clusters.
Han says NetApp is looking to take Kubernetes to another level to address what in some instances already involves petabytes of data. Not all that data may need to be accessed on a local cluster, but he notes organizations most likely will have fleets of Kubernetes clusters trying to access a common pool of data. What will be required is a way to manage data on Kubernetes clusters that is more application-centric, he says.
NetApp has already made significant investments in Kubernetes, including NetApp Trident, an open source controller for provisioning storage in container environments, and NetApp Kubernetes Services, a control plane for creating and managing Kubernetes clusters.
Kubernetes, of course, is accelerating adoption of hyperconverged infrastructure (HCI) platforms both inside and out of the cloud. Most cloud platforms are based on HCI platforms, an architecture that unifies compute and storage resources and is now being employed more widely in local data centers as well. Kubernetes will then provide a common layer of abstraction that can be applied consistently across local data centers and multiple public clouds. However, Han notes this vision of hybrid cloud computing will not be realized if IT organizations are still locked into specific databases and data management platforms.
It may be some time before Project Astra makes it possible to achieve that goal. In fact, it’s not clear to what degree platform providers might even resist making data truly portable across platforms. However, as Kubernetes becomes more widely employed, it’s only a matter of time before more attention is paid to data management. Project Astra may not be the final word on the subject, but it does provide a place to start what is arguably a long-overdue conversation.