Changes and New Features History

NVIDIA UFM Cyber-AI Documentation v2.5.0



Cyber-AI plugins

Added Running Cyber-AI Plugin

Deploying UFM Cyber-AI from an ISO File

Added instructions on deploying UFM Cyber-AI from an ISO file. For more information, refer to Appendix - Deploying UFM Cyber-AI from an ISO File



DPU telemetry Integration

Accumulate telemetry data from ethtool and sysfs providers.

Cyber-AI Appliance OS Remanufacture

Added instructions on how to remanufacture the Cyber-AI appliance and OS. For more information, refer to Appendix - Cyber-AI Appliance OS Remanufacture.



Morpheus integration

Included Morpheus Integration in the production level

Cable Anomalities Detection

Added additional properties to the Cable Animalities Detection table

Cable Analysis Improvements

Added the following:

  • Option to filter by clicking on the bars

  • Added percentage to bars

Automatic Evaluation License

Generated default evaluation license on first launch

Reorganized Cyber AI Tools

All Cyber AI tools are documented and have the ‘ufm-cai-‘ prefix



Recommended Actions

Recommended actions for anomalies and alerts were improved to give a recommendation procedure and steps to follow to fix this alert/anomaly

Support SLURM based on UFM Logical-Servers

Aggregate data from devices that belongs to the same logical server, analyze this data and find alerts or anomalies at the logical server level

Combining of Cable info into one tab

Two cable tabs were combined into one tab for better user experience

Filtering Up to Down only for Anomaly View

SanKey graphs are not to be filtered once other objects filtered

Refresh button adding to all UFM Cyber-AI tabs

Added manual and automatic refresh per each dashboard

Adding Version Number to every tab in UFM Cyber-AI

Version number should be available on every tab in UFM Cyber-AI

Detecting incompatible FW version in UFM Cyber-AI

Sometime the statistics are coming zero due to incompatible version of FW.
The comparison of several parameters such as RX_power <>0, TX_bias=0 and Link_Up=true will provide with recommended action to upgrade the software.

Morpheus integration

Morpheus Integration was tested in Beta level, but not included

Integrate GPU usage for model training

Using GPU to enhance performance of model training was tested in POC level, but not included

Integration and Infrastructure improvement

  • Integrated latest versions of UFM Telemetry and UFM Enterprise

  • Improved scheduler settings infrastructure



Cable Anomalies

Added new cable anomalies analysis based on cable attributes trend, a tachometer indication was added also for each anomaly

HA Service (2 nodes)

Added high-availability (HA) support for two Cyber-AI appliances based on DRBD and Pacemaker

Weekly average

Added ability to display weekly average graphs for relevant counters

Data cleanup

Added support for data cleanup, purge, or archiving of old UFM Cyber-AI data files

Anomalies Analysis View

Added new tab for Anomaly Analysis view

© Copyright 2023, NVIDIA. Last updated on Sep 11, 2023.