Physical Resource Overview

"Physical Resource Overview" is designed for platform administrators to aggregate compute resource costs from three dimensions — physical nodes, GPU card types, and individual cards — and configure physical resource billing strategies to support cluster operational cost accounting.

Entry path: Left sidebar Billing > Physical Resource Overview, page route /admin/billing/machine.

Page Menu

Page	Route	Description
Node Resources	/admin/billing/machine/node-resource	Physical node cost details and Top analysis
Card Type Resources	/admin/billing/machine/node-type-resource	GPU vendor/card model cost analysis
GPU Card Details	/admin/billing/machine/card-resource	Individual GPU card cost details
Resource Billing Strategy	/admin/billing/machine/strategy	CPU/memory/GPU physical resource pricing

Node Resources

View Node Costs

1. Navigate to Billing > Physical Resource Overview > Node Resources.
2. The top of the page displays the cluster cost overview: total cost, CPU cost, memory cost, GPU compute cost, and GPU memory cost.
3. Below is the node cost detail table: index, name, IP, CPU cost, memory cost, GPU memory cost, GPU compute cost, total cost.
4. Use the time range selector at the top to choose the statistics period.
5. Use the search box to search by node name (fuzzy) or filter by IP (exact); both the list and statistics update in real time.

Top Node Analysis

1. The middle of the page displays two pie charts:
   • Top 5 Node Memory Cost Pie Chart
   • Top 5 Node Compute Cost Pie Chart
2. Hover over a sector to view the specific node and its cost proportion.

Export Data

1. Click "Export CSV" in the upper-right corner of the list.
2. The system downloads the node cost details under the current filter conditions.

Card Type Resources

View Vendor/Card Model Costs

1. Navigate to Billing > Physical Resource Overview > Card Type Resources.
2. The top of the page displays a vendor cost pie chart (aggregated by GPU vendor) and a Top 5 card type cost pie chart.
3. Below is the card type cost detail table: index, card type name, vendor, GPU memory cost, GPU compute cost, total cost.
4. Use the "Vendor" dropdown at the top to filter the card type list; custom vendor names not in the preset list can be entered manually.

Drill Down to GPU Card Details

1. In the card type cost detail table, click "View Details" in a row.
2. The page navigates to the "GPU Card Details" page with the card type and vendor filters pre-filled automatically.

GPU Card Details

View Individual Card Costs

1. Navigate to Billing > Physical Resource Overview > GPU Card Details.
2. Use the top filter bar to query by node, vendor, and GPU card model; the three dimensions are cascading — after selecting a vendor, the card model options update automatically.
3. The list displays: index, GPU card alias/ID, parent node, GPU card model, vendor, GPU memory cost, GPU compute cost, total cost.
4. Custom vendor or model input is supported.

Export Data

1. After setting the filter conditions, click "Export CSV".
2. The system downloads the GPU card cost details under the current filter conditions.

Resource Billing Strategy

View Billing Strategy

1. Navigate to Billing > Physical Resource Overview > Resource Billing Strategy.
2. The list displays: rule name, resource type, GPU vendor, card model, billing price, billing cycle, application scope, and rule description.
3. Use the top filter bar to query by rule name and resource type.

Create Billing Strategy

1. Click "Create Strategy".
2. Fill in:
   • Rule Name: An easily identifiable name.
   • Resource Type: Machine CPU / Machine Memory / GPU Memory / GPU Compute.
   • GPU Vendor + Card Model: Cascading selection when the resource type is GPU-related; locked during editing.
   • Billing Price: Unit price.
   • Billing Cycle: Hour / Day / Month, etc.
   • Application Scope: Platform-wide or a specific workspace.
   • Rule Description: Optional.
3. Click "Save".

Manage Billing Strategy

1. Click "More Actions" in the list row:
   • Edit: Modify price, cycle, application scope, etc.; resource type and GPU vendor/card model are locked during editing.
   • Delete: Remove a strategy that is no longer in use.

FAQ

Q: What is the difference between physical resource cost and tenant resource cost? A: Physical resource cost aggregates operational costs from the cluster's physical nodes/cards perspective; tenant resource cost aggregates resource consumption and cost allocation from the workspace/project/user perspective.

Q: Why do some nodes show no GPU cost? A: Only nodes that have actually been allocated GPU resources will incur GPU compute/memory costs; pure CPU nodes display only CPU and memory costs.

Q: How long after modifying a billing strategy does it take effect? A: Changes typically take effect immediately; historical costs already incurred are not retroactively adjusted, and only new billing cycles after the modification are affected.