Research Notes

Can Portworx Unify Data Management for VMs, Containers, and AI on Kubernetes?

Research Finder

Find by Keyword

Can Portworx Unify Data Management for VMs, Containers, and AI on Kubernetes?

Kube Datastore, AI data services, and expanded observability extend enterprise storage capabilities as customers move virtualization and production AI workloads onto Kubernetes.

11/18/2026

Key Highlights

  • Portworx by Everpure announced general availability of Kube Datastore (KDS) in Portworx Enterprise 3.7.2 for FlashArray storage and Red Hat OpenShift. Dynamic pools use native SAN resilience and performance capabilities, with Portworx managing Kubernetes workload placement and failover. Static pools retain replication-based availability for workloads that require it. Support for third-party SAN arrays and additional Kubernetes platforms is planned for 2027.
  • Pre-staged volumes, targeted for Q1 2027, will accelerate AI inference startup by distributing local copies of model weights among Kubernetes nodes. Portworx also supports data locality, hybrid-cloud mobility, and policy-based storage management throughout the AI lifecycle, including protection for datasets, model artifacts, and intermediate work product.
  • Portworx expanded observability through its Operations Center, added Red Hat OpenShift 5 support, and advanced Virtbench 2.0, an open-source benchmarking framework for Kubernetes virtualization. The company reports more than 75 customers running Kubernetes virtualization in production, with approximately 51,000 VMs and 190,000 storage volumes deployed.

The News

Portworx by Everpure announced new Kubernetes data management capabilities at KubeCon North America 2026. Kube Datastore is now generally available for FlashArray and Red Hat OpenShift, allowing Kubernetes environments to use native SAN capabilities for VM storage and recovery. Forthcoming pre-staged volumes target AI inference startup delays, while expanded observability, Red Hat OpenShift 5 support, and virtualization benchmarking extend the platform's capabilities. The announcements advance Portworx's strategy to provide a common data management layer for virtual machines, containers, and AI workloads. For more information, see the official company announcement.

Analyst Take

Portworx is extending its Kubernetes data management architecture as enterprises move virtual machines and AI applications onto a common platform. Virtual machines bring established expectations for storage performance and recovery. AI introduces shared datasets, data locality requirements, and changing GPU capacity. Portworx's challenge is to give administrators consistent data services while preserving established storage infrastructure.

Kube Datastore changes how Portworx uses enterprise SAN infrastructure. Traditional Portworx deployments provide resilience through software-managed replication among Kubernetes nodes. When the underlying array already provides redundancy, data reduction, and controller failover, additional replication consumes capacity and network bandwidth. KDS Dynamic Pools use these array capabilities while managing Kubernetes-level storage placement and recovery. Customers can reduce replication overhead while continuing to use their existing SAN arrays.

The HyperFRAME Research Lens: State of the AI Stack (3Q 2026) found that only 15% of organizations have fully modernized, AI-ready data architectures, while 52% identified scalability and performance limitations as barriers to data platform modernization. KDS allows customers to extend existing SAN investments into Kubernetes virtualization while modernizing their data architecture in stages.

Bringing enterprise storage capabilities into Kubernetes virtualization

KDS Dynamic Pools allow shared SAN storage to be presented through Kubernetes nodes and reassigned when workloads move or nodes fail. Because the data remains on the underlying array, recovery avoids rebuilding software-managed replicas. Portworx also consolidates persistent volumes within managed pools, reducing the number of individual storage objects provisioned on the array. Platform engineers can manage these functions through Kubernetes workflows while the array supplies data reduction, redundancy, and storage performance.

Portworx reports up to two to three times faster recovery in certain failure scenarios, with additional performance gains from FastPath for single-replica volumes. The architecture also looks to introduce compute-aware placement, allowing workloads and storage pools to be distributed among available nodes. Portworx reports approximately 51,000 VMs running among more than 75 virtualization customers, compared with roughly 45 customers six months earlier. One global manufacturing customer reportedly completed a VMware migration involving more than 50 manufacturing sites, 80 data centers, and 150 Kubernetes clusters, using Portworx to manage storage among multiple backends.

The initial urgency following VMware's licensing changes has moderated, but enterprises continue evaluating alternative virtualization platforms. Customers still need to establish whether replacement platforms can meet their existing application and recovery requirements. Familiar datastore concepts and integration with OpenShift Virtualization can help customers transition while preserving established service levels.

Kubernetes pod status alone does not establish when a guest operating system is ready to serve applications. Virtbench measures VM boot storms, live migration, failure recovery, storage I/O, and time to guest readiness, giving administrators reproducible tests for evaluating infrastructure before production cutover. Portworx is working to incorporate the open-source tool into the KubeVirt project.

Competition intensifies for Kubernetes data management

Portworx competes with Red Hat OpenShift Data Foundation, NetApp Trident and Trident Protect, Dell Container Storage Modules, and Kubernetes-native storage platforms such as SUSE Longhorn. NetApp integrates ONTAP storage and protection capabilities with OpenShift Virtualization, while Dell connects its enterprise arrays to Kubernetes through CSI drivers and management integrations. Red Hat OpenShift Data Foundation provides software-defined storage integrated with OpenShift.

Kube Datastore gives Portworx a differentiated approach by combining established SAN capabilities with Kubernetes-native storage management. The wider Portworx platform supports multiple Kubernetes distributions and storage backends, although KDS Dynamic Pools initially support FlashArray and OpenShift. Third-party SAN integration will be important to extending that differentiation.

Competition also comes from integrated infrastructure platforms such as Nutanix, which combines virtualization, Kubernetes management, and persistent storage services. Portworx has an opportunity among enterprises seeking flexibility in their infrastructure choices, particularly those maintaining heterogeneous storage environments.

Extending data management into the AI lifecycle

AI storage requirements change throughout model development and deployment. Training datasets may need shared read access among GPU nodes, while vector databases and retrieval-augmented generation indexes require persistent storage, consistent snapshots, and recovery. Platform teams can define Kubernetes storage classes for different AI workloads, automating provisioning while enforcing placement, performance, and protection policies. Portworx can also preserve notebooks, pipelines, evaluation sets, model artifacts, and scheduling logic so interrupted work can resume without recreating these assets.

The ROSEN Group, an industrial inspection company, uses Red Hat OpenShift and Kubeflow to process pipeline inspection data. Portworx reports that ROSEN automated approximately 99% of its platform provisioning processes and achieved 70% faster safety analysis. Portworx also supports hybrid-cloud mobility, enabling organizations to place workloads near their data and use temporary cloud GPU capacity when needed.

Pre-staged volumes extend these capabilities into inference. The forthcoming feature downloads model weights once and distributes local copies to eligible Kubernetes nodes, reducing repeated transfers as GPU workloads scale or restart. Storage policies govern placement and access to these copies. Portworx has identified precomputed execution graphs for compatible vLLM workers as another potential use case. Supporting persistent agent memory and changing inference state would require additional consistency and synchronization capabilities.

What Was Announced

Portworx announced general availability of Kube Datastore in Enterprise 3.7.2 for FlashArray and Red Hat OpenShift, supporting up to 12 Dynamic Pools and 700 volumes per node. The release adds compute-aware placement and FastPath, while static pools retain replication-based availability. Third-party SAN and additional Kubernetes platform support are planned for 2027.

Virtualization enhancements include FlashArray XCOPY migration, reportedly up to 12 times faster than the default path, and more granular VM disaster recovery selection. Virtbench 2.0 adds VM lifecycle, migration, and recovery benchmarks, with Red Hat proposing its use for OpenShift Virtualization CSI certification.

Pre-staged volumes, targeted for Q1 2027, will distribute local model-weight copies to accelerate AI inference. Portworx also supports hybrid-cloud AI data mobility, including temporary cloud GPU use and return of results and metadata.

Operations Center adds pool inventory, migration progress, topology-based troubleshooting, and consolidated Portworx and Prometheus alerting with impact visibility. Portworx also announced OpenShift 5 support. Portworx Backup 3.2.0 adds Windows VM selective file restore and IPv6 support, while planned Enterprise 3.8.0 enhancements will enable third-party backup using Kubernetes changed block tracking (KEP-3314).

Looking Ahead

The planned expansion of KDS to third-party SAN arrays will test how consistently Portworx can support storage placement, mobility, and recovery in heterogeneous environments. FlashArray integration establishes the initial implementation, but other arrays have different capabilities and performance characteristics. Portworx will need to demonstrate the reported recovery improvements under production conditions. Virtbench could provide a framework for comparing VM lifecycle performance as customers evaluate Kubernetes virtualization.

Backup integration and multicluster management also need further development. Customers moving from VMware will expect established protection and recovery processes to work as applications transition to Kubernetes. The planned changed block tracking support in Enterprise 3.8.0 should expand third-party backup options, while Operations Center needs to extend its troubleshooting and management capabilities across multiple clusters.

For AI, we expect Portworx to bring pre-staged volumes to market in Q1 2027 and demonstrate measurable improvements in inference startup time and GPU utilization. The underlying technology also presents opportunities to accelerate other parts of the inference process, including reusable execution graphs and potentially additional forms of application state. As these capabilities develop, Portworx will need to account for data consistency and synchronization, particularly when information changes during execution. Hybrid-cloud mobility offers another area for expansion, allowing enterprises to use GPU capacity where it is available while maintaining control over application data, models, and associated metadata. Production deployments should help establish how effectively these capabilities support AI workloads that move between on-premises and cloud environments.

Author Information

Don Gentile | Analyst-in-Residence, Data Platforms & Resiliency

Don Gentile brings three decades of experience turning complex enterprise technologies into clear, differentiated narratives that drive competitive relevance and market leadership. He has helped shape iconic infrastructure platforms including IBM z16 and z17 mainframes, HPE ProLiant servers, and HPE GreenLake — guiding strategies that connect technology innovation with customer needs and fast-moving market dynamics. 

His current focus spans flash storage, storage area networking, hyperconverged infrastructure (HCI), software-defined storage (SDS), hybrid cloud storage, Ceph/open source, cyber resiliency, and emerging models for integrating AI workloads across storage and compute. By applying deep knowledge of infrastructure technologies with proven skills in positioning, content strategy, and thought leadership, Don helps vendors sharpen their story, differentiate their offerings, and achieve stronger competitive standing across business, media, and technical audiences.