The Topology view maps your network infrastructure in a hierarchical, real-time display. Filter by job ID to isolate the spine, leaf, and compute nodes carrying traffic for a specific workload, then overlay anomaly data to identify which components are affecting performance. The topology includes a dedicated hierarchy for NVLink and NVSwitch fabric alongside the Ethernet view.
Interface Components
|
Component |
Description |
|---|---|
|
Filter Bar |
Filter by time range (for example, "Current Day"), job ID, device and interface role, anomaly and event display, Device-to-Device Path, Flow Pattern, or Job Topology. |
|
Visual Legend |
Defines the color-coded node types: Dc (Datacenter), Pd (Pod), Ssp (Super Spine), Sp (Spine), Lf (Leaf), and server/compute nodes. |
|
Job Details Panel |
Shows metadata for the selected job: name (for example, |
|
Show Controls |
Overlay anomalies, events, and metrics on the topology. Select Metric to display metric values on links; select Anomalies or Events to add operational context to the visualization. |
|
View Selector |
Switch between saved topology views (for example, |
Features
Multi-Layer Hierarchy
The topology organizes your network into nested levels you can expand and collapse:
-
Dc (Datacenter): top-level container
-
Pd (Pod): logical grouping of network resources
-
Ssp (Super Spine): aggregation-layer switches connecting spine layers across pods
-
Sp (Spine): high-capacity backbone switches
-
Lf (Leaf): access-layer switches connecting directly to compute nodes
-
Server/Node: compute hosts at the leaf level
-
NIC: host network interface cards
Click any link in the topology to open the link detail window. This window shows an events table listing recent events on that link, and metrics including current and historical values. Use the link detail window to investigate a specific connection without leaving the topology view.
Navigation
Expand and collapse hierarchy levels by clicking the node label. Use the hierarchy toggle to switch between hierarchical and non-hierarchical (flat) views. Zoom and pan the canvas using the mouse scroll wheel and click-drag, or use the on-screen zoom controls.
Overlays
Filter the topology by a specific job ID (for example, 4335424). This isolates the spine (Sp), leaf (Lf), and node devices participating in that job, removing all unrelated cluster topology. This allows you to view which switches and links are carrying traffic for a failing or degraded job. You can alternately view link metrics through the metrics overlay.
Integrated Anomaly Mapping
NVIDIA Minerva overlays performance telemetry directly onto topology nodes, so you can visually correlate job performance with network health without switching views. Common overlays include:
-
-
Anomalous PFC (Priority Flow Control): visualizes where pause frames are being generated, indicating downstream congestion
-
RoCE retransmits: highlights links or adapters experiencing packet loss that triggers RDMA retransmission, often a precursor to job timeouts
-
Rare log anomalies: surfaces unique log patterns via NVIDIA Minerva's ML engine
-
Save and Share Topology Views
Save and share topology configurations using named views.
To create a new view:
-
Configure the filters, overlays, and hierarchy settings you want to save.
-
Select +Add in the view selector.
-
Enter a name for the view and select Publish to make it available to other users.
To switch between views, select the view name from the view dropdown (for example, topology_v1).
Maintenance Mode
Place a device in Maintenance Mode during scheduled maintenance windows to avoid false alerts and skewed health metrics. When a device is in Maintenance Mode, NVIDIA Minerva does not send notifications or events for that device.
Put a Device in Maintenance Mode
-
Go to Topology > Provision > Metadata > Devices and locate the target device.
-
Click
-
Set Admin Status to Maintenance.
To disable Maintenance Mode and resume alerts, set Admin Status back to Enabled.
Last updated: