Integrated Dell Remote Access Controller 9 User's Guide

Data Processing Unit (DPU)

A Data Processing Unit (DPU) is a system on a chip that consists of ARM cores, a NIC ASIC, and acceleration engines. A DPU is programmable and potentially capable of running an operating system. DPU support is added to iDRAC from release 6.00.30.00 onwards. DPUs combine network connectivity with CPU cores independent from the Hypervisor or operating system, allowing acceleration and offload services. DPUs distinguish themselves from traditional offload engines by their flexibility, programmability, and ability to host a wide variety of services.

Use of DPU offers the following advantages:

• Isolates infrastructure services from the host operating system and applications.
• Enables an environment to deliver new services independent of the host application environment.
• Enables hardware acceleration to perform data-intensive operations at wire-speed.
• Free up server/x86 CPU cores for enabling customer applications single socket and small form factor edge platforms

After the DPU operating system is booted, additional PCIe Functions can be initialized. So, the BIOS PCIe enumeration (and host Operating System/Hypervisor boot process) shall happen only after DPU operating system has booted or ready.

iDRAC allows you to configure the DPU operating system Ready Mode (Boot Synchronization) settings against each DPU capable slot. Possible values are:

• Enabled: DPU does participate in holding the BIOS PCIe enumeration and host Operating System/Hypervisor boot process.
• Disabled: DPU does not participate in holding the BIOS PCIe enumeration and host Operating System/Hypervisor boot process.

Points to consider about DPU:

• Only a few slots are DPU capable. iDRAC allows you to configure DPU Boot Synchronization against those slots alone.
• DPU Boot Synchronization settings are Slot-based (Not Identity based). That is, if the DPU device is moved to a different slot, the device behaves as per the newly inserted Slot configuration.
• DPU Boot Synchronization settings are configurable even without the presence of a DPU device.
• After discovery, if the slot does not have a DPU device that is installed, then DPU Boot Synchronization configurations are NOT effective.
• Individual DPU operating system Ready and Overall DPU operating system Ready are reported in the LCL.
• On an unsupported DPU platform, while performing a system erase, the DPU LC logs display the SYS560 message (some of the DPU devices failed to reset). On a supported DPU platform, if the DPU is not present and a system erase is performed, the logs display the SYS564 message (unable to perform system erase of DPU because there is no DPU available in the system).
• When the NVIDIA BF3 cards are disabled from the BIOS configuration, the Status in the Network Devices page. (System > Overview > Network Devices > Summary) in iDRAC UI is displayed in Green.
• When the PCIe slots of NVIDIA BF3 DPU cards are Boot Driver Disabled in the BIOS configuration (System BIOS Settings > Integrated Devices > Slot Disablement), PR7 and PR8 logs are displayed in the iDRAC LC logs.
• The device hardware inventory is displayed for the NVIDIA BF3 cards when the PCIe slot is Boot Drive Disabled from the BIOS configuration.
• After you perform the powercycle operation and set the powersavingmode ON on the Nokia card, the BIOS may stop and the Non-Maskable Interrupt (NMI) message is displayed. In this case, reboot the system.
• When the Nokia RAN DPU card is set to the powersaving mode, critical LC logs are displayed.

Following are the features of DPU:

• You are allowed to configure DPU Boot Synchronization for each DPU capable slot.
• You are allowed to configure the DPU operating system Ready timeout value in minutes (from 0 to 30).
• Based on the user configuration, BIOS PCIe enumeration and host Operating System/Hypervisor boot process happens only after each Boot synchronization enabled DPU has reported that DPU operating system is ready.
• Other PCIe functions that are exposed by DPU operating system are enumerated by BIOS and reported in the iDRAC Hardware Inventory.
• BIOS displays various DPU-related messages during the POST:
  • Discovering Data Processing Unit(s) ...—While discovering the DPU devices
  • Discovering Data Processing Unit(s) ... Done—When DPU discovery is completed.
  • Initializing Data Processing Unit (Do NOT reboot the system)—At 0, 10, 20, 30, 40, 50, 60, 70, 80, 90, and 100% completion of boot-synchronization
  • Initializing Data Processing Unit... Done—When 100% completion of boot-synchronization and boot-synchronization is successful.
• Individual and overall DPU operating System Ready messages are reported in LCL.
  NOTE:DPU operating system Ready flag persists across the Host reboots, and the DPU operating system Ready message is logged for each Host reboot.
• If any LC-SSM task is present, then BIOS skips the waiting on DPU boot synchronization.

Inventory and Monitoring of DPUs

iDRAC system inventory provides the make and model of the DPU while monitoring the health of the DPU cores, peripherals, and the installed operating system. GET is used to retrieve inventory information. This action ensures that no unauthorized devices are installed maliciously. Using the GET operation, you can periodically check the DPU health. If the system is healthy, it returns a payload response and status update to provide health updates.

To detect malicious or accidental DPU operating system installations use GET operation. Using the GET operation that you can retrieve operating system name, vendor name, version, and status of the DPU operating system.

You can view the installed DPU from iDRAC UI (System > Inventory > Firmware) Inventory

Redfish

Using Redfish, you can set a one-time boot configuration which is used to boot the DPU with the configured value once on reboot. On the next reboot, the DPU boot is based on the Boot Order that is configured. Redfish also allows you ARM-UEFI and BMC firmware updates. For more information, see developer.dell.com.

Serial Console

To access serial control using RACADM, login to iDRAC SSH - Racadm> console dpu

Co-ordinated Shutdown

The ESXi operating system shutdown process internally shutdown the DPU ESXio to protect ESXio File corruption.