Skip to main content

Dashboard

Overview​

The dashboard is the main page for system administrators to get an at-a-glance view of the overall status of the GPU management system. It provides visualizations of real-time GPU usage, project-specific allocations, and request statuses using cards, charts, and tables.

①~③ Card Section – Summary of Key Metrics (Total GPUs Card, Utilization Card, and Pending Requests Card)

④ Usage Analysis Chart – GPU Usage Analysis

⑤ Top GPU Pools Table – List of Main GPU Pools

⑥ Pending GPU Requests Table – List of Pending Requests

Card Section​

Total GPUs Card​

  • Purpose: Display the total number of GPU resources registered in the system.
  • Color: Theme default.
  • Displayed information:
    • Total GPUs: Total number of GPUs across the infrastructure.
    • GPU count by status: GPU counts for each status: Ready, Failed, and Mismatched.
  • Action: Clicking the card navigates to the GPU Pools Management page for further details.
  • Use cases:
    • Provide a quick overview of infrastructure scale, including a breakdown of operational status.
    • Serve as a reference for utilization calculations.

Utilization Card​

  • Purpose: Displays overall GPU utilization (%) in real-time by 'Available' and 'In Use'.
  • Color: Green tones.
  • Displayed information:
    • Utilization: Percentage of GPUs currently in use.
    • Available: Number of GPUs ready for allocation.
    • In Used: Number of GPUs currently allocated and running workloads.
  • Action: Clicking the card navigates to the GPU Capacity Management page for a deeper dive into usage analytics.
  • Use cases:
    • Enable users to monitor resource consumption
    • Quickly identify available capacity and determine if additional resources are needed.
info

Utilization = (In Used / Total GPUs) × 100%

Pending Requests Card​

  • Purpose: Displays the number of GPU requests pending approval.
  • Color: Yellow tones.
  • Displayed information:
    • Pending Requests: Count of requests with status REQUESTED.
    • Icon: Checkbox icon to signal pending actions.
  • Action: Clicking the card navigates to the GPU Pool Requests page with the REQUESTED filter applied.
  • Use case: Instantly monitor unprocessed requests and prioritize actions accordingly.
warning

A high number of pending requests may require immediate attention to avoid delays in resource allocation.

Usage Analytics Chart​

Cluster Usage Chart​

  • Purpose: Visualize the allocation status of GPUs by specification using a polar donut chart, delivering instant insights into cluster-wide resource distribution.
  • Color: Green tones.
  • Display information:
    • Total: The total GPU count for the selected cluster.
    • Allocated: Number of GPUs assigned to pools but not currently in use (green).
    • Available: Number of GPUs not yet allocated and available for new assignments (light green).
    • Used: Number of GPUs currently being utilized by running pods (gray).
  • Use cases:
    • Compare allocation, usage, and availability across GPU specs.
    • Identify usage patterns and resource bottlenecks.
    • Support decision-making for scaling and resource planning.
  • Interpretation logic:
    • Allocated: Represents GPUs allocated to pools but not yet used.
    • Used: Represents GPUs actively used by workloads (usedProfile from allCounts API).
    • Available: Represents GPUs that are neither allocated nor in use (Available from cluster-capacity).
    • Allocated + Used: Represents all GPUs allocated to pools (from cluster-capacity).
    • Chart slice size reflects actual numbers for each status/spec.

Usage by Project Chart​

  • Purpose: Shows the percentage of GPU usage by project. You can visually see how much each project accounts for in the total usage.
  • Color: Purple tones.
  • Display information:
    • Top 5 projects are shown individually with number of GPUs in use.
    • The other projects are grouped under "Other"
    • Total: Total number of GPUs in use by all projects
  • Use cases:
    • Monitor which projects are consuming the most GPU resources.
    • Detect concentration and usage trends.
    • Track project-specific changes over time.
  • Interpretation logic: Slice size = usage amount per project.
note

"Other" cannot be expanded; details are available via tooltip.

Usage by GPU Type Chart​

  • Purpose: Illustrate GPU usage by hardware type with a donut chart for quick reference (A100, H100, etc.).
  • Color: Blue tones.
  • Display information:
    • Top 5 types shown individually with number of GPUs in use.
    • The other types are grouped under "Other".
    • Total: Total GPUs in use by type.
  • Use cases:
    • Identify the most and least utilized GPU types.
    • Detect type-based availability and saturation.
    • Analyze hardware trends and inform future procurement.
  • Interpretation logic: Slice size = usage count per GPU type.

Top 5 GPU Pools Table​

  • Purpose: Provide a real-time overview of the main GPU pools, highlighting usage, assignment, and status to support effective resource management.
  • Display information:
    • Pool Name: Name of the GPU pool, clickable to navigate to the pool’s details page.
    • Project: Assigned project name; if unassigned, displayed as “Unassigned”.
    • Type: GPU model and quantity information (e.g., 7g.40gb x1, 1g.5gb x1, Mixed).
    • Usage: Shows Used/Total GPUs (e.g., 0/1, 1/6), visualized with a progress bar.
    • Status: Pool Status (Healthy/Unhealthy)
      • Inuse: Active pool, currently in use.
      • Available: Pool is available for assignment.
      • Unassigned: No project assigned.
  • Features:
    • Sorting: Sort by Pool Name and Usage.
    • View All: Access the complete GPU pool list.
    • Usage: Usage is indicated by progress bar color.
      • Low Usage: Green color.
      • Medium Usage: Yellow color.
      • High Usage: Red color.
  • Use case: Monitor usage and status of key pools, identify unassigned pools and plan allocations.
    • Monitor the usage rate and operational status of key GPU pools.
    • Identify “Unassigned” pools for future allocation planning.
    • Quickly review pool health and prioritize management actions.

Pending GPU Requests Table​

  • Purpose: Display the top 5 most recent pending GPU Pool requests, allowing administrators to efficiently manage approval workflows.
  • Display information:
    • Project: Name of the project in which the request was created
    • Pool Name: Name of the requested GPU pool.
    • Requested Date: Date and time when the request was created.
    • Action: Approve or Reject the request (allows the administrator to Approve or Reject the request).
  • Features:
    • Sorting: Sort by Pool Name or Requested Date.
    • Actions:
      • Approve: Instantly approve the request.
      • Reject: Opens a modal to input the rejection reason (mandatory)
    • View All: Access the complete list of GPU Pool requests
  • Use Case:
    • Quickly identify and process older or urgent requests.
    • Streamline approval and rejection, minimizing administrative delays.
  • Processing Method:
    1. Click Approve → request is approved
    2. Click Reject → input reason in modal → request is rejected
note
  • Data updates automatically after processing.
  • Rejection requires a reason to ensure transparency and traceability.