| id | system-config |
|---|
This tool is to collect the system information automatically on the tested GPU nodes including the following hardware categories:
-
Install SuperBench on the local machine using root privilege.
-
Start to collect the sys info using
sb node info --output-dir ${output-dir}command using root privilege. -
After the command finished, you can find the output system info json file
sys_info.jsonof local node under ${output_dir}.
-
Install SuperBench on the local machine.
-
Deploy SuperBench onto the remote machines.
-
Prepare the host file of the tested GPU nodes using Ansible Inventory on the local machine.
-
After installing the Superbnech and the host file is ready, you can start to collect the sys info automatically using
sb run --get-infocommand. The detailed command can be found from SuperBench CLI.
sb run --get-info -f host.ini --output-dir ${output-dir} -C superbench.enable=none
- After the command finished, you can find the output system info json file
sys_info.jsonof each node under ${output_dir}/nodes/${node_name}.
| SubCategory | Key | Command | Description | Example |
| OS | system-manufacturer | dmidecode -s system-manufacturer | manufacturer of the system | Microsoft Corporation |
| system-product name(virtual machine) | dmidecode -s system-product-name | product name or virtual machine | Virtual Machine | |
| operating_system | cat /proc/version | version of current running os | Ubuntu 9.3.0-17ubuntu1~20.04 | |
| uname | uname | short for system information | Linux sb-test-wu-000000 5.8.0-1039-azure #42~20.04.1-Ubuntu | |
| Docker | docker_server_version | docker version | server version of docker engine | 20.10.3 |
| docker_client_version | docker version | client version of docker engine | 20.10.3 | |
| VM | vmbus | lsvmbus | devices attached to the Hyper-V VMBus |
"VMBUS ID 1": "[Dynamic Memory]", "VMBUS ID 2": "Synthetic mouse", ... |
| Kernel | kernel_modules | lsmod | list of active kernel modules |
"Module": "binfmt_misc", "Size": "24576", "Used": "1" ... |
| kernel_parameters | sysctl | kernel parameters |
"abi.vsyscall32": "1", "debug.exception-trace": "1", ... |
|
| DMI | dmidecode | dmidecode | DMI table dump (info on hardware components) | "dmidecode": "# dmidecode 3.2\nGetting SMBIOS data from sysfs..." |
| SubCategory | Key | Command | Description | Example |
| General | model | dmidecode -t memory | distinct model name of the memory | Samsung M393A4K40DB3-CWE |
| type | dmidecode -t memory | distinct type of memory | DDR4-3200 | |
| clock frequency | dmidecode -t memory | distinct clock frequency of memory | 3200 MT/s | |
| channels | dmidecode -t memory | the number of memory chips | 16 | |
| capacity | lsmem | the total capacity of memory | 511.9G | |
| block_size | lsmem | the block size of memory | 128M |
| SubCategory | Key | Command | Description | Example |
| General | archeticture | lscpu | architecture of cpu | x86_64 |
| model name | lscpu | model name of cpu | AMD EPYC 7662 64-Core Processor | |
| cpu op-mode | lscpu | cpu mode: 32bit/64bit | 32-bit, 64-bit | |
| byte order | lscpu | byte order | Little Endian | |
| address size | lscpu | size of address | 48 bits physical, 48 bits virtual | |
| cpus | lscpu | logical cpu cores count | 256 | |
| On-line CPU(s) list | lscpu | on-line logical cpu cores | 0-255 | |
| Thread(s) per core | lscpu | thread per core | 2 | |
| Core(s) per socket | lscpu | core per socket | 64 | |
| Socket(s) | lscpu | socket count | 2 | |
| NUMA node(s) | lscpu | numa node count | 4 | |
| L<x> caches | lscpu | cache size | "L1d cache": "4 MiB", "L1i cache": "4 MiB", "L2 cache": "64 MiB", "L3 cache": "512 MiB" | |
| NUMA node<x> CPU(s) | lscpu | cpu core list of the numa node | "NUMA node0 CPU(s)": "0-31,128-159", "NUMA node1 CPU(s)": "32-63,160-191", "NUMA node2 CPU(s)": "64-95,192-223", "NUMA node3 CPU(s)": "96-127,224-255" | |
| Flags | lscpu | cpu flags | fpu vme de pse tsc msr pae mce cx8 apic ... | |
| max_speed | sudo dmidecode -t processor | grep "Speed" | distinct cpu max frequency | 3700 MHz | |
| current_speed | sudo dmidecode -t processor | grep "Speed" | distinct cpu current frequency | 2000 MHz |
| SubCategory | Key | Command | Description | Example |
| FileSystem | filesystem | df -Th | the name/path of the filesystem | /dev/nvme0n1p2 |
| avail | df -Th | avail size of the filesystem | 1.4T | |
| size | df -Th | total size of the filesystem | 1.8T | |
| type | df -Th | the type of the filesystem | ext4 | |
| block_size | blockdev --getbsz /dev/<device> | the block size of the filesytem | 4096 | |
| 4k_alignment | 4kDEVICE=/dev/sdb1 do parted $DEVICE align-check opt 1; done_alignment | whether the file system is 4k alignment | 1 aligned | |
| BlockDevice | name | lsblk -e 7 -o NAME,ROTA,SIZE,MODEL | the name of the block device | nvme0n1 |
| model | lsblk -e 7 -o NAME,ROTA,SIZE,MODEL | the model name of the block device | VO001920KXAVP | |
| rotational | lsblk -e 7 -o NAME,ROTA,SIZE,MODEL | whether rotational, thai is HDD or SSD | 0 | |
| size | lsblk -e 7 -o NAME,ROTA,SIZE,MODEL | the total size of the block device | 1.8T | |
| block_size | fdisk -l -u /dev/{} | grep "Sector size" | the sector size of the block device | Sector size (logical/physical): 512 bytes / 512 bytes | |
| General | mapping | mount | mount relationship between filesystem and block device |
| SubCategory | Key | Command | Description | Example |
| NIC | nic_logical_name | lshw -c network | logical name of the nic | ib1 |
| nic_model | lshw -c network | model name of the nic | Mellanox Technologies MT28908 Family [ConnectX-6] | |
| nic_firmware | lshw -c network | fw version | 20.30.1004 (MT_0000000594) | |
| nic_driver | lshw -c network | driver version | mlx5_core[ib_ipoib] 5.3-1.0.0 | |
| nic_speed | lshw -c network | speed spec of the nic | 200 Gbit/s | |
| nic_disabled | lshw -c network | whether diabled | false | |
| IB | device_info | ibv_devinfo -v | list of device information for each ib device | "hca_id:\tmlx5_0": ... |
| device_status | ibstat | list of device status for each ib device | "CA 'mlx5_0'": ... | |
| General | ofed_version | ofed_info -s | the version of ofed | MLNX_OFED_LINUX-5.3-1.0.5.0: |
| SubCategory | Key | Command | Description | Example(NVIDIA) | Example(AMD) |
| General | driver_version | nvidia-smi -q -x/rocm-smi -a | driver version | 460.27.04 | 5.9.25 |
| topology | nvidia-smi topo -m/rocm-smi --showtopo | gpu connection topology (nvidia only) | / | / | |
| nvidia-container-runtime_version | nvidia-container-runtime -v | version of nvidia-container-runtime (nvidia only) | 1.0.0-rc92 | N/A | |
| nvidia-fabricmanager_version | nv-fabricmanager --version | version of nvidia-fabricmanager (nvidia only) | 460.27.04 | N/A | |
| nv_peer_mem_version | dpkg -l | grep 'nvidia-peer-memory' | version of nv_peer_mem (nvidia only) | 1.1-0 | N/A | |
| GPUCard | rocm_info | rocm-smi -a & rocm-smi --showmeminfo vram | amd gpu info of each gpu&lsindex>, including firmware, frequency, memory, etc. (amd only) | N/A | "card0": ... "card1": ... |
| nvidia_info | nvidia-smi -q | nvidia gpu info list of each gpu, including firmware, frequency, memory, etc. (nvidia only) |
"timestamp": "Fri Aug 20 05:36:24 2021", "driver_version": "460.27.04", "cuda_version": "11.2", "attached_gpus": "8", "gpu": [...] ... |
N/A |
| SubCategory | Key | Command | Description | Example |
| General | topology | lspci -t -vvv | topology of installed PCI devices | / |
| device_info | lspci -vvv | device info on installed PCI devices | 00:00.0 Host bridge: Advanced Micro Devices, Inc. [AMD] Starship/Matisse Root Complex... |