Nvidia BlueField-2 BF2500 User Manual
Nvidia BlueField-2 BF2500 User Manual

Nvidia BlueField-2 BF2500 User Manual

Ethernet dpu controller
Hide thumbs Also See for BlueField-2 BF2500:
Table of Contents

Advertisement

Quick Links

NVIDIA BlueField-2 BF2500 Ethernet DPU Controller User Manual
Exported on Oct/31/2023 01:37 PM

Advertisement

Table of Contents
loading
Need help?

Need help?

Do you have a question about the BlueField-2 BF2500 and is the answer not in the manual?

Questions and answers

Subscribe to Our Youtube Channel

Summary of Contents for Nvidia BlueField-2 BF2500

  • Page 1 NVIDIA BlueField-2 BF2500 Ethernet DPU Controller User Manual Exported on Oct/31/2023 01:37 PM...
  • Page 2: Table Of Contents

    Table of Contents Introduction............................9 System Requirements ............................9 Package Contents ...............................10 Card Package ..............................10 Accessories Kit .............................. 10 Product Overview ...............................11 Ideal Solution for JBOF and JBOD Systems......................11 Features and Benefits ............................11 Supported Interfaces ..........................14 Layout and Interfaces ............................14 Interfaces Detailed Description..........................17 DPU................................
  • Page 3 1GbE OOB Management Interface ........................20 1GbE OOB Management LEDs Interface......................21 RTC Battery ..............................21 eMMC Interface.............................. 21 External PCIe Power Supply Connector........................21 Networking Ports LEDs Interface ......................... 22 Pin Description ............................23 PCI Express Interface............................23 Power Sequencing............................28 NC-SI Management Interface ..........................28 External Power Supply Connector (6-pin ATX Connector)....................35 Hardware Installation ..........................
  • Page 4 Cables and Modules.............................39 Networking Cables ............................39 RJ45 Cable for the 1GbE OOB Management Interface....................40 USB Interface Cabling ............................41 Bring-Up and Driver Installation ........................42 BlueField Software..............................42 Software On eMMC..............................43 Yocto Distribution Installation ..........................43 PXE Server Configuration on Host Side ........................44 Installing Linux on BlueField Controller Card.......................44 Software Requirements ............................
  • Page 5 Installing MLNX_OFED on the BlueField Controller Card ..................62 Updating BlueField Controller Card Firmware ..................... 64 Troubleshooting ............................66 Specifications............................67 MBF2H516B-CEEOT / MBF2H516B-CENOT Specifications....................67 BlueField-2 BF2500 Mechanical Drawing and Dimensions....................68 Finding the MAC on the Controller Card......................70 Thermal Sensors............................71 Document Revision History ......................... 72  ...
  • Page 6 About This Manual It provides details as to the interfaces of the board, specifications, required software and firmware for operating the board, and a step-by-step plan of how to bring up the BlueField-2 DPU. EOL'ed (End of Life) Ordering Part Numbers Series/ Core Speed Max Speed No.
  • Page 7 Copper cables (DACs), copper splitter cables, Active Optical Cables (AOCs) and transceivers in a wide range of lengths from 0.5m to 10km. In addition to meeting IBTA standards, NVIDIA tests every product in an end-to-end environment ensuring a Bit Error Rate of less than 1E-15. ...
  • Page 8 Revision History A list of the changes made to this document are provided in Document Revision History.
  • Page 9: Introduction

    This is the User Manual for NVIDIA® BlueField®-2 BF2500 DPU Controller. This document provides details of the product interfaces, specifications, required software and firmware for operating the board, and a step-by-step plan of how to bring up the BlueField-2 BF2500 DPU Controller.  System Requirements Item Description Main-board PCI Express slot...
  • Page 10: Package Contents

    Package Contents Before installing your new system, unpack it and check against the below tables that all the parts have been sent. Check the parts for visible damage that may have occurred during shipping.  If anything is damaged or missing, contact your reseller. Card Package ...
  • Page 11: Product Overview

    Ideal Solution for JBOF and JBOD Systems  NVIDIA BlueField-2 DPU is a highly integrated and efficient controller, optimized for NVMe storage systems, Network Functions Virtualization (NFV), Cloud and Machine Learning workloads. BlueField-2 integrates all the discrete components of a storage system appliance into a single chip, including Arm core CPUs, PCIe switch and a network controller, making it the premier solution for building Just-a-Bunch-Of-Flash (JBOF) systems, All-Flash-Array and storage appliances for NVMe over Fabrics.
  • Page 12 NVIDIA PeerDirect NVIDIA PeerDirect communication provides high-efficiency RDMA access by eliminating unnecessary internal data copies between components on the PCIe bus (for example, from GPU to CPU), and therefore significantly reduces application run time. DPU advanced acceleration technology enables higher cluster...
  • Page 13 Feature Description Quality of Service (QoS) Support for port-based Quality of Service enabling various application requirements for latency and SLA. Storage Acceleration A consolidated compute and storage network achieves significant cost-performance advantages over multi-fabric networks. Standard block and file access protocols can leverage RDMA for high-performance storage access.
  • Page 14: Supported Interfaces

    Supported Interfaces Layout and Interfaces This section describes the DPU supported interfaces. Each numbered interface that is referenced in the figures is described in the following table with a link to detailed information.  The below figures are for illustration purposes only and might not reflect the current revision of the BF2500 card. Component Side...
  • Page 15 Print Side...
  • Page 17: Interfaces Detailed Description

    Interface x8 NAND flash Interfaces Detailed Description NVIDIA® BlueField®-2 DPU is a family of advanced DPU IC solutions that integrate a coherent mesh of 64-bit Arm v8 A72 cores, an NVIDIA® ConnectX®-6 Dx network adapter front-end and a PCI Express switch into a single chip. The powerful DPU IC architecture includes an Arm v8 multicore processor array and enables customers to develop sophisticated applications and highly differentiated feature sets.
  • Page 18: Encryption

    At the heart BlueField-2, the ConnectX-6 Dx network offload controller with RDMA and RDMA over Converged Ethernet (RoCE) technology delivers cutting- edge performance for networking and storage applications such as NVMe over Fabrics. Advanced features include an embedded virtual switch with programmable access lists (ACLs), transport offloads and stateless encaps/decaps of NVGRE, VXLAN, and MPLS overlay protocols. Encryption ...
  • Page 19: Ddr4 Sdram On-Board Memory

    DDR4 SDRAM On-Board Memory  The DPU incorporates 16GB @ 3200MT/s single DDR4 channel, 64bit + 8bit ECC, solder-down memory. NC-SI Management Interface  The DPU enables the connection of a Baseboard Management Controller (BMC) to a set of Network Interface Controller (NICs) for the purpose of enabling out-of-band remote manageability.
  • Page 20: Usb Interfaces

     It is prohibited to directly connect any RS-232 cable! Only TTL 3.3V voltage level cables are supported.  The USB to UART cable is not used for NC-SI management purposes. USB Interfaces The controllers use a mini-USB Type B connector to load operating system images. 1GbE OOB Management Interface The DPU incorporates a 1GbE RJ45 out-of-band port that allows the network operator to establish trust boundaries in accessing the management function to apply it to network resources.
  • Page 21: 1Gbe Oob Management Leds Interface

    1GbE OOB Management LEDs Interface There are 2 OOB management LEDs, one green and one amber/yellow. The following table describes LED behavior for DPUs with or with on-board BMC. LED Indications Link Activity Green LED Amber/Yellow LED Link off Link on / No activity Blinking 1 Gb/s link / Activity (RX,TX) Other combinations...
  • Page 22: Networking Ports Leds Interface

    The FHHL P-Series DPUs with x16 PCIe Gen 4 lanes incorporate an external 12V power connection through a 6-pin ATX connector. The DPU includes a special circuitry that provides current balancing between the two power supplies; the 12V from the PCIe x16 standard slot and the 12V from the ATX 6-pin connector.
  • Page 23: Pin Description

    Pin Description PCI Express Interface The BF2500 Controller Card uses a PCI Express x16 connector (component and print sides) according to the PCI Express 4.0 specifications. The following two tables provide component side pins description and print side pins description.  Pin # Signal Name Description Pin #...
  • Page 24 Ground RSVD Not connected REFCLK+ Output PCIe Reference Differential Clock. Not connected (default) Ground REFCLK- Output PCIe Reference Differential Clock. Not connected (default) PETP0 Input DPU SoC RX Ground PETN0 Input DPU SoC RX PERP0 Output DPU SoC TX Ground PERN0 Output DPU SoC TX...
  • Page 25 PERP3 Output DPU SoC TX Ground PERN3 Output DPU SoC TX RSVD Not connected Ground RSVD Not connected RSVD Not connected Ground RSVD Not connected PETP4 Input DPU SoC RX Ground PETN4 Input DPU SoC RX PERP4 Output DPU SoC TX Ground PERN4 Output...
  • Page 26 Ground PETN7 Input DPU SoC RX PERP7 Output DPU SoC TX Ground PERN7 Output DPU SoC TX RSVD Not connected Ground Ground RSVD Not connected PETP8 Input DPU SoC RX Ground PETN8 Input DPU SoC RX PERP8 Output DPU SoC TX Ground PERN8 Output...
  • Page 27 Ground PETN11 Input DPU SoC RX PERP11 Output DPU SoC TX Ground PERN11 Output DPU SoC TX Ground Ground PETP12 Input DPU SoC RX Ground PETN12 Input DPU SoC RX PERP12 Output DPU SoC TX Ground PERN12 Output DPU SoC TX Ground Ground PETP13...
  • Page 28: Power Sequencing

    PERP15 Output DPU SoC TX Ground PERN15 Output DPU SoC TX PRSNT2# Output x16 PCIe card detection, connected to PRESNT1# Ground RSVD Not connected Power Sequencing The BF2500 Controller Card requires the 12V ATX PCIe power to come up prior or at the same time to the +12V at the PCIe edge connector. NC-SI Management Interface...
  • Page 33 NC ...
  • Page 34 NC  ta  ta ...
  • Page 35: External Power Supply Connector (6-Pin Atx Connector)

    External Power Supply Connector (6-pin ATX Connector) The below table provides the External Power Supply pins of the external power supply interfaces on the DPU. For further details, please refer to External PCIe Power Supply Connector. The mechanical pinout of the 6-pin external +12V power connector is shown below. The +12V connector is a GPU power PCIe standard connector. Care should be taken to ensure the power is applied to the correct pins as some 6-pin ATX type connector can have different pinouts.
  • Page 36: Hardware Installation

    Hardware Installation Installation and initialization of the BlueField-2 BF2500 DPU Controller require attention to the mechanical, power, and precautions for rack-mounted equipment. Safety Warnings  Safety warnings are provided here in the English language. For safety warnings in other languages, refer to the...
  • Page 37: System Requirements

    Hardware Requirements  Unless otherwise specified, NVIDIA products are designed to work in an environmentally controlled data center with low levels of gaseous and dust (particulate) contamination. The operating environment should meet severity level G1 as per ISA 71.04 for gaseous contamination and ISO 14644-1 class 8 for cleanliness level.
  • Page 38: Software Requirements

    Software Requirements • See Operating System under System Requirements section under the Introduction section. • Software Stacks - The BF2500 DPU Controller is shipped with Linux based Operating System burned on it which includes all needed drivers. For more information, please refer to the BlueField-2 Software User Manual. •...
  • Page 39: Installation Instructions

    This section provides detailed instructions on how to install your BlueField-2 2500 DPU Controller in a system.  The BlueField-2 BF2500 DPU Controller should be installed only in a JBOF System as it functions as a PCIe root-complex (RC) initiating PCIe bus operations. Installing it in a regular host system may damage the card.
  • Page 40: Rj45 Cable For The 1Gbe Oob Management Interface

    Insert the connector into the adapter card. Be careful to insert the connector straight into the cage. Do not apply any torque, up or down, to the connector cage in the adapter card. Verify that the connector locks in place.   When installing cables make sure that the latches engage.
  • Page 41: Usb Interface Cabling

    USB Interface Cabling Debugging and loading new versions of the operating systems and firmware requires the use of Type A to Type B mini-USB 2.0 Cable. The following figure shows an example of a cable with a USB Type A connector on one end an a mini-USB Type B connector on the other.
  • Page 42: Bring-Up And Driver Installation

    BlueField SW ships with the NVIDIA BlueField Controller Cards. BlueField SW is a reference Linux distribution based on the Yocto Poky distribution and extended to include the NVIDIA OFED stack for Arm and a Linux kernel which supports NVMe-oF. This SW distribution is capable of running all customer- based Linux applications seamlessly.
  • Page 43: Software On Emmc

     Before installing the preferred OS on the BlueField Controller Card, make sure you install the card only in a JBOF system. Installing it in a host system may damage the card. Software On eMMC The BlueField Controller Card boots off eMMC upon power-up. The image flashed on the eMMC from the factory is the Yocto Linux. Run the following command to discover the BlueField Software version: cat /etc/bluefield_version Yocto Distribution Installation...
  • Page 44: Pxe Server Configuration On Host Side

    shutdown -r now Verify the version via: cat /etc/bluefield_version PXE Server Configuration on Host Side Before installing CentOS 7 on the BlueField Controller Card, you need to configure the PXE server on the host side (x86) to allow the deployment of the CentOS image over the BlueField Controller Card.
  • Page 45: Software Requirements

     Please note that CentOS 7.5 is not supported. • Access to the latest BlueField Controller Card SW bundle: NVIDIA uses box.com to distribute BlueField software. Contact your sales/support representative for a custom link to download BlueField software releases. •...
  • Page 46: Setup Procedure With Installation Script

    The “-i” flag points to the OS installation disk. This is the image that is accessed via PXE boot to install the OS on the BlueField Controller Card. • The “-o” flag points to the NVIDIA OFED driver disk for Arm. Download and extract it from  http://www.mellanox.com/page/products_dyn? product_family=34.
  • Page 47 To install the kernel modules, please follow the instruction in section RShim Host Driver. To install the kernel modules, please follow the instruction in NVIDIA BlueField DPU Family Software Documentation under BlueField Software Overview > Installation and Initialization > RShim Host Driver. Step 3: Configure TFTP Server The host should be configured to act as a TFTP server to the BlueField Controller Card via the USB RShim network.
  • Page 48 Patch the initrd with the eMMC driver and TMFIFO (RShim network) driver: mkdir -p /tmp/.bfcentos mkdir -p $/tmp/.bfinstdd cd /tmp/.bfcentos xzcat /var/lib/tftpboot/centos/7.4/initrd-orig.img | cpio -idm mount /root/BlueField-1.0.alpha3.10409/distro/rhel/bluefield_dd/bluefield_dd-4.11.0- 22.el7a.aarch64.iso /tmp/.bfinstddmkdir -p usr/lib/modules/4.11.0-22.el7a.aarch64/updates/cp /tmp/.bfinstdd/lib/modules/4.11.0-22.el7a.aarch64/updates/dw_mmc*.ko usr/lib/mod- ules/ 4.11.0-22.el7a.aarch64/updates/cp /tmp/.bfinstdd/lib/modules/4.11.0-22.el7a.aarch64/updates/tmfifo.ko usr/lib/mod- ules/4.11.0-22.el7a.aarch64/ updates/cp /root/BlueField-1.0.alpha3.10409/distro/rhel/bluefield_dd/bluefield_dd-4.11.0- 22.el7a.aarch64.iso ./ bluefield_dd.iso umount /tmp/.bfinstdd;...
  • Page 49 systemctl restart tftp  Based on the system, the user may need to use “system TFTP restart” instead. Also, if required, the user might need to switch use “xinetd” instead of “TFTP”. Step 4: Set Up the DHCP Server DHCP server set up on the host is required for BlueField Controller Card to get a private IP from the host for PXE boot process completion. Configure the correct server names and domain names so that the BlueField Controller Card can connect to the network via the host later on.
  • Page 50: Flashing Bluefield Controller Card Bootloader Code

    # Specify the IP address for this client. host pxe_client { hardware ethernet 00:1a:ca:ff:ff:01; fixed-address 192.168.100.2; } EOF  It is recommended to back up the previous dhcpd.conf file before overwriting it. Step 5: Set Up the HTTP Server The TFTP server allows the PXE boot to load the initrd and kernel. The BlueField Controller Card obtains all the other required sources through the network, thus, making it necessary to set up an HTTP.
  • Page 51: Opening Terminal Connection To Bluefield Controller Card

    Opening Terminal Connection to BlueField Controller Card To open a console window to the BlueField Controller Card, a terminal application is required. The application “minicom” is used for the flow, however, any standard terminal application can work, e.g. “screen”.  Install minicom by running “yum install minicom”...
  • Page 52 [root@bu-lab02 ~]# ls /dev/rshim0/ boot console net rshim The boot device is used to push the bootstream to the BlueField Controller Card. Upon writing to it, it automatically resets the Arm cores so that it is booted using the pushed bootstream. ...
  • Page 53: Installing Centos 7.4 On Bluefield Controller Card

    Installing CentOS 7.4 on BlueField Controller Card  If the error “no root is found” appears in the installation process, check or disable the firewall as needed on the server host machine. Full PXE Boot Installation Get to the UEFI boot menu. Reboot the BlueField Controller Card by typing “reboot”...
  • Page 54 Navigate to the Boot Manager. UEFI Boot Manager  Select EFI Network, it will then use the TFTP service on the host to discover all available PXE boot options. Shortly after, a “..Fetching Netboot Image” message will appear enabling CentOS installation.  Option to Install CentOS Select CentOS download.
  • Page 55  These configuration inputs are not needed when the kickstart option “-k” is specified when running the setup.sh script. ====================================================================================== ====================================================================================== VNC   Text mode provides a limited set of installation options. It does not offer custom partitioning for full control over the disk layout.
  • Page 56 Europe Pacific 10) Arctic Asia Australia 11) US America 8) Atlantic 12) Etc Africa Indian Antarctica   Please select the timezone. Use numbers or type names directly [b to region list, q to quit]: 11 ====================================================================================== ====================================================================================== Timezone settings   Available timezones in region US Alaska 4) Eastern...
  • Page 57 [ ] File and Print Server 9) [ ] Development and Creative [ ] Basic Web Server Workstation Please make your choice from above ['q' to quit | 'c' to continue | 'r' to refresh]: 9 ====================================================================================== ====================================================================================== Base environment Software selection  ...
  • Page 58 [ ] 3) Use Free Space   Installation requires partitioning of your hard drive. Select what space to use for the install target.   Please make your choice from above ['q' to quit | 'c' to continue | 'r' to refresh]: c ====================================================================================== ====================================================================================== Partition Scheme Options [ ] 1) Standard Partition [ ] 2) Btrfs...
  • Page 59 ====================================================================================== ====================================================================================== Please select new root password. You will have to type it twice.   Password: Password (confirm): ====================================================================================== ====================================================================================== Question   The password you have provided is weak: The password fails the dictionary check it is based on a dictionary word. Would you like to use it anyway?  ...
  • Page 60: Non-Pxe Boot Installation

    Non-PXE Boot Installation When the setup script is run with the “-t” option, it generates a nonpxe.bfb file at the directory where the script is run. The directory contains the install kernel and rootfs which are usually loaded by UEFI during the initial PXE boot stage. Thus, if pushing this file, the host TFTP server no longer needs to be used and UEFI would automatically load the install kernel and rootfs from the boot FIFO.
  • Page 61: Installing Mlnx_Ofed On Arm Cores

    Installing MLNX_OFED on Arm Cores Prerequisite Packages for Installing MLNX_OFED • MLNX_OFED installation requires some prerequisite packages to be installed on the system. Currently, CentOS installed on the BlueField Controller Card has a private network to the host via the USB connection, and it can be used to Secure Copy Protocol (SCP) all the required packages.
  • Page 62: Removing Pre-Installed Kernel Module

    yum install rpm-build yum group install "Development Tools" yum install kernel-devel-`uname -r` yum install valgrind-devel libnl3-devel python-devel yum install tcl tk Note that this is not needed if you installed CentOS 7 with the kickstart (“-k”) option. /auto/sw_mc_soc_project/distro/rhel/kernel-devel-4.11.0-22.el7a.aarch64.rpm Removing Pre-installed Kernel Module There are cases where the kernel is shipped with an earlier version of the mlx5_core driver taken from the upstream Linux code.
  • Page 63 mount /root/MLNX_OFED_LINUX-4.2-1.4.8.0-rhel7.4alternate-aarch64.iso /mnt Install MLNX_OFED. If the kernel on the BlueField is 4.11.0-22.el7a.aarch64, run: cd /mnt # ./mlnxofedinstall --bluefield If the kernel is different than 4.11.0-22.el7a.aarch64, run: cd /mnt # ./mlnxofedinstall --add-kernel-support --skip-repo  For OFED to support DPDK, use the arguments “--upstream-libs” and “--dpdk”. ./mlnxofedinstall --distro rhel7.4alternate --add-kernel-support --upstream-libs --dpdk This step might take longer than expected to be completed.
  • Page 64: Updating Bluefield Controller Card Firmware

    Updating BlueField Controller Card Firmware  The below commands apply to MBF1M616A-CSNAT. The commands vary per OPN. The below steps demonstrate how to manually update the firmware if the automatic process fails. The firmware image can be found in the BlueField Software package.
  • Page 65 To burn the firmware which comes with OFED after OFED is installed, run: /opt/mellanox/mlnx-fw-updater/firmware/mlxfwmanager_sriov_dis -force Power cycle the BlueField Controller Card for the new firmware to take effect. root@bluefield:~# flint -i /opt/fw-Bluefield-red-18_99_4608_MBF1M6X6A-CSNA_Ax- Flexboot-3.5.404_UEFI-14.15.20.bin -d /dev/mst/mt41682_pciconf0 b   Current FW version flash: 18.24.0013 New FW version: 18.99.4608...
  • Page 66: Troubleshooting

    Troubleshooting • Reseat the BF2500 Controller Card in its slot or a different slot, if necessary The BF2500 Controller Card no longer works • Try using another cable • Reboot the system • Try removing and re-installing all BF2500 Controller Cards BF2500 Controller Card stopped working after installing another BF2500 card •...
  • Page 67: Specifications

    Single-channel with 8 DDR4 8 bit + ECC (64bit + 8bit ECC) 16GB @ 3200MT/s On-board Memory 64GB eMMC memory Voltage: 12V DPU Power Supply Power and airflow specifications are provided in NVIDIA BlueField-2 DPUs Power and Airflow Specifications document, Power Consumption and Air Flow which is available at NVOnline following login.
  • Page 68: Bluefield-2 Bf2500 Mechanical Drawing And Dimensions

    CE / FCC / VCCI / ICES / RCM RoHS RoHS compliant a. The non-operational storage temperature specifications apply to the product without its package. BlueField-2 BF2500 Mechanical Drawing and Dimensions  All dimensions are in millimeters. The PCB mechanical tolerance is +/- 0.13mm.
  • Page 69 Bracket Mechanical Drawing and Dimensions...
  • Page 70: Finding The Mac On The Controller Card

    Finding the MAC on the Controller Card Each DPU Controller has a different identifier printed on the label: serial number and the card MAC for the Ethernet protocol.  The product revisions indicated on the labels in the following figures do not necessarily represent the latest revisions of the cards. Board Label (Example)  ...
  • Page 71: Thermal Sensors

    Thermal Sensors The BlueField-2 BF2500 DPU Controller incorporates the BlueField-2 DPU device which operates in the range of temperatures between 0°C and 105°C. There are three thermal threshold definitions for the BlueField-2 DPU which impact the overall system operation state: •...
  • Page 72: Document Revision History

    Document Revision History Date Comments/Changes May. 2023 Updated Specifications - added non-operational storage temperature specifications Jun. 2021 • Removed OPNs: MBF2H515B-VEEOT, MBF2H515B-VENOT • Multiple text and drawing updates throughout the manual Dec. 2020 • Updated the NCSI interface pinouts. • Removed the UART interface from "Supported Interfaces". •...
  • Page 73 NVIDIA product and may result in additional or different conditions and/or requirements beyond those contained in this document. NVIDIA accepts no liability related to any default, damage, costs, or problem which may be based on or attributable to: (i) the use of the NVIDIA product in any manner that is contrary to this document or (ii) customer product designs.
  • Page 74 INDIRECT, SPECIAL, INCIDENTAL, PUNITIVE, OR CONSEQUENTIAL DAMAGES, HOWEVER CAUSED AND REGARDLESS OF THE THEORY OF LIABILITY, ARISING OUT OF ANY USE OF THIS DOCUMENT, EVEN IF NVIDIA HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. Notwithstanding any damages that customer might incur for any reason whatsoever, NVIDIA’s aggregate and cumulative liability towards customer for the products described herein shall be limited in...

This manual is also suitable for:

Mbf2h516b-ceeotMbf2h516b-cenot

Table of Contents