Aviz ONES and VAST Data Enable Scalable, Multi-Tenant AI Data Platform

Power the backbone of your AI Factory with secure, high-performance, multi-tenant data infrastructure.

March 13, 2026

Download PDF

As cloud service providers and enterprises build large-scale AI and HPC environments, performance, security, and scalability of a multi-tenant, unified data platform environment become crucial. Modern AI factories increasingly operate as shared platforms, supporting multiple teams, workloads, and tenants on common infrastructure. While high-performance storage is essential to keep GPUs consistently fed with data, enterprises must also ensure secure isolation, predictable networking, and operational simplicity as environments scale.

VAST Data delivers a high-performance, Disaggregated Shared Everything (DASE) architecture purpose-built for AI and HPC workloads, enabling organizations to scale storage throughput and capacity independently. VAST’s AI OS is a unified data platform supporting File/Object/Block protocols, database, and compute engines into a single, linearly scalable platform. The VAST networking architecture is purpose-built to enable this model, providing exabyte-scale capacity and performance, secure isolation.

Aviz ONES (Open Networking Enterprise Suite) provides centralized orchestration and NetOps to connect GPU, compute, and storage resources securely. Together, Aviz and VAST help CSPs and enterprises deploy a foundational AI Operating System with multitenancy, greater consistency, control, and efficiency for entire AI and HPC environments.

Figure 1: Aviz ONES and Vast Data Storage Integration

Aviz ONES enables intent-based design and automated provisioning of the network fabric connecting the VAST Data Platform to compute and GPU clusters. Through policy-driven workflows and per-tenant segmentation, ONES ensures that multiple tenants can securely share the same high-performance storage and unified, intelligent foundation for intelligent, autonomous AI systems, without compromising isolation or predictability. This approach simplifies Day-0 design and deployment, Day-1 tenant management, and Day-2 network monitoring while reducing the risk of configuration errors in complex AI environments.

Beyond initial deployment, ONES addresses the operational challenges of running shared AI infrastructure at scale. Built-in visibility across devices and fabrics provides operators with real-time insight into storage connectivity and network health. Lifecycle automation capabilities, including drift detection, configuration compliance, and expansion workflows, help teams manage growth and change without manual intervention.

By combining VAST Data’s scalable AI Data platform with Aviz ONES’ intelligent orchestration and lifecycle automation, CSPs and enterprises can move beyond siloed, manually operated environments. The joint solution provides a more predictable, secure, and operationally efficient foundation for multi-tenant AI infrastructure, enabling organizations to build and scale AI workloads with confidence and simplicity while gaining control over operational complexity.

Aviz ONES provides two options to manage tenant isolation and segmentation in the N-S network.

Tenant isolation is achieved using either:

1. VLAN/VRF-based segmentation (Tenant aware)
2. VXLAN/VRF-based segmentation (Tenant unaware)

In addition, the user can decide whether to have an end-to-end tenant network segmentation across N-S (user & storage) network or choose a common storage segment for all tenants.

Tenant Un-aware Storage Network

In a tenant-unaware model, tenant isolation is applied only to the user-facing (North–South) network, while the storage network remains shared across all tenants. The storage fabric operates within a single default VRF, independent of tenant boundaries, and tenant access is enabled through route leaking mechanisms. This model simplifies design and deployment by maintaining a common storage domain. It is best suited for environments with a limited number of tenants (typically up to ~20) where operational simplicity is prioritized over strict isolation within the storage layer.

Tenant-Aware Storage Network

In a tenant-aware model, tenant isolation is extended into the storage fabric using VRF-based segmentation. Each tenant maintains a distinct logical boundary across compute, access, and storage networks, ensuring consistent policy enforcement and isolation throughout the infrastructure.

This approach enables a more scalable and policy-consistent architecture, making it suitable for environments that require large-scale multi-tenancy (hundreds to thousands of tenants).

CSP large scale and isolated Multi-Tenancy

The VAST cluster network isolation approaches above allow subdividing the Server Pools for creating isolated domains, making it possible to provision the performance of a required size pool of VIPs to a set of users or applications to isolate application traffic and ensure a quality of ingress and egress performance (QoS) that’s not possible in shared-nothing or shared-disk architectures.

1. Control Access to Data with network-based protection to isolate data sets to designated tenants​.
2. Performance Isolation to eliminate “Noisy Neighbor” issues​.
3. Optimize resource utilization by scaling of pools via API, adding and removing servers to meet changing needs.