InfiniBand Fabric Managed by UFM
NVIDIA®UFM is a host-based solution that provides all the management functionalities required for managing fabrics.
Fabric Topology with UFM
UFM Server is a server on which UFM is installed and has complete visibility over the fabric to manage routing on all devices.
UFM HA Server is a UFM installed server on a secondary server for High Availability deployment.
Managed Switching Devices are fabric switches, gateways, and routers managed by UFM.
Managed Servers are the compute nodes in the fabric on which the various applications are running, and UFM manages all servers connected to the fabric.
UFM Host Agent is an optional component that can be installed on the Managed Servers. UFM Host Agent provides local host data and host device management functionality.
The UFM Host Agent provides the following functionality:
- Discovery of IP address, CPU, and memory parameters on host
- Collection of CPU/Memory/Disk performance statistics on host
- Upgrading HCA Firmware and OFED remotely
- Creating an IP interface on top of the InfiniBand partition
UFM Switch Agent is an embedded component in NVIDIA switches that allows IP address discovery on the switch and allows UFM to communicate with the switch. For more information, please refer to Device Management Feature Support.