Provisioning

Instead of provisioning being handled separately through scripts, vendor tools and manual configuration, ClusterVisor provides a centralized platform where provisioning is standardized, repeatable and visible end to end.

One interface for the entire provisioning lifecycle

ClusterVisor brings all provisioning activities into a single GUI-driven control plane, including:

• Compute node provisioning and integration
• GPU resource enablement and allocation
• Storage mounting and configuration
• Network and cluster fabric setup validation
• Scheduler integration and queue configuration
• User and project environment setup

This eliminates the need to coordinate across multiple disconnected tools or administrative domains.

Hardware-aware provisioning

Because ClusterVisor is tightly integrated with ACT-engineered systems, provisioning is hardware-aware from the start. It ensures newly added resources are:

• Automatically discovered in the cluster
• Validated for health and readiness
• Properly categorized (CPU, GPU, storage tiers and so on)
• Integrated into monitoring and telemetry
• Available to scheduling and workload systems

This closes the “bring-up gap” where hardware is installed but not yet usable.

Policy-driven configuration

ClusterVisor provisions based on defined policies rather than manual configuration steps. For example:

• GPU nodes can be automatically assigned to AI queues
• High-memory nodes can be grouped for simulation workloads
• Storage tiers can be mapped to project requirements
• User groups can be linked to predefined resource allocations

This keeps the environment consistent and reduces configuration drift over time.

Integrated user and project provisioning

Unified provisioning extends beyond hardware into the user layer. ClusterVisor can align with identity systems and internal policies to provision:

• User accounts and project spaces
• Access permissions based on roles or groups
• Resource quotas for teams or departments
• Job submission environments and defaults

A new user or research group can be fully operational much faster.

Scheduler and workload readiness

Provisioning isn’t complete until the cluster can actually run work. ClusterVisor ensures:

• Job schedulers are configured and synchronized
• Queues reflect available hardware
• Policies match organizational priorities
• Resource limits and priorities are correctly enforced

Provisioned resources are immediately usable, not just technically online.

Automation through plugins

ClusterVisor’s plugin architecture lets provisioning workflows be extended and automated, for example:

• Auto-registration of new hardware
• Integration with external IT or cloud systems
• Custom environment setup for specific applications
• AI- or simulation-ready cluster templates
• Environment cloning for research projects

Reduced time to productivity

Unified provisioning shortens the time between hardware delivery and a usable HPC environment. By standardizing and automating provisioning, ClusterVisor helps organizations:

• Deploy clusters faster
• Reduce manual configuration errors
• Stay consistent across expansions
• Scale AI and HPC workloads rapidly

ClusterVisor acts as a single orchestration layer for the entire HPC stack, coordinating hardware, software, users and workloads through a centralized, policy-driven and extensible platform.