Building commands

In order to properly build commands for Intel® Tiber™ Broadcast Suite’s image, it is needed to gather information about the platform used for execution, and properly put those into the command.

Note: This instruction bases the examples on a configuration of one of the machines, the responses may differ.

NIC-related settings

This chapter explains how to gather the platform information regarding network card/s and Virtual Functions, and use them in commands.

Container’s IP address

Use the network defined for docker containers to determine a pool of addresses (this may also be taken from output of first_run.sh).

docker network ls

Response example:

NETWORK ID     NAME           DRIVER    SCOPE
d25f79b8d31d   bridge         bridge    local
b334bae7126d   host           host      local
b49202de451f   my_net_801f0   bridge    local
c06348a2888f   none           null      local

my_net_801f0 network is created by first_run.sh.

Then determine the Subnet based on the docker’s network ID or name:

docker network inspect my_net_801f0 | grep Subnet

or

docker network inspect b49202de451f | grep Subnet

Response example:

                    "Subnet": "192.168.2.0/24",

IP address should be taken from the subnet 192.168.2.0/24.

Chosen IP address must be used in --ip=x.x.x.x as part of [docker_parameters], as well as -p_sip x.x.x.x part of [broadcast_suite_parameters].

Example: --ip=192.168.2.2 and matching -p_sip 192.168.2.2.

Note: [docker_parameters] is used for every flag that stands between docker run and an image name (e.g. tiber_broadcast_suite). [broadcast_suite_parameters] determines every switch that is put after the image name.

Find a proper E810 Series NIC

Run lshw command for network interfaces and find all E810 Series cards.

sudo lshw -C network

Response example:

(...)
  *-network:0
       description: Ethernet interface
       product: Ethernet Controller E810-C for QSFP
       vendor: Intel Corporation
       physical id: 0
       bus info: pci@0000:c0:00.0
       logical name: eth0
       version: 02
       serial: 6c:fe:54:5a:18:70
       capacity: 25Gbit/s
       width: 64 bits
       clock: 33MHz
       capabilities: pm msi msix pciexpress vpd bus_master cap_list rom ethernet physical 25000bt-fd autonegotiation
       configuration: autonegotiation=off broadcast=yes driver=ice driverversion=Kahawai_1.13.7_20240220 firmware=4.40 0x8001c967 1.3534.0 latency=0 link=no multicast=yes
       resources: iomemory:2eff0-2efef iomemory:2eff0-2efef irq:16 memory:2efffa000000-2efffbffffff memory:2efffe010000-2efffe01ffff memory:f1600000-f16fffff memory:2efffd000000-2efffdffffff memory:2efffe220000-2efffe41ffff
  *-network:1
       description: Ethernet interface
       product: Ethernet Controller E810-C for QSFP
       vendor: Intel Corporation
       physical id: 0.1
       bus info: pci@0000:c0:00.1
       logical name: eth1
       version: 02
       serial: 6c:fe:54:5a:18:71
       capacity: 25Gbit/s
       width: 64 bits
       clock: 33MHz
       capabilities: pm msi msix pciexpress vpd bus_master cap_list rom ethernet physical fibre 25000bt-fd autonegotiation
       configuration: autonegotiation=on broadcast=yes driver=ice driverversion=Kahawai_1.13.7_20240220 duplex=full firmware=4.40 0x8001c967 1.3534.0 latency=0 link=yes multicast=yes
       resources: iomemory:2eff0-2efef iomemory:2eff0-2efef irq:16 memory:2efff8000000-2efff9ffffff memory:2efffe000000-2efffe00ffff memory:f1500000-f15fffff memory:2efffc000000-2efffcffffff memory:2efffe020000-2efffe21ffff
(...)

Check which card/s have link=yes in configuration part. Above, only eth1 fulfills this requirement. A card with link=no should not be used.

Note the following information for all of the matching cards:

• bus info

• logical name

• serial

• configuration

Virtual Function’s port address

Knowing the PCI address of the proper NIC (example: eth1), use the command below to determine addresses of Virtual Functions.

for vf in /sys/bus/pci/devices/<PHYSICAL_DEVICE_PCI_ADDRESS>/virtfn*
do
  basename $(readlink -f "$vf")
done

Note: Physical address of device is indicated as Slot in lspci’s response, or bus info in lshw’s response.

Example:

for vf in /sys/bus/pci/devices/0000:c0:00.1/virtfn*
do
  basename $(readlink -f "$vf")
done

0000:c0:11.0
0000:c0:11.1
0000:c0:11.2
0000:c0:11.3
0000:c0:11.4
0000:c0:11.5

Based on above response, Virtual Functions derived from eth1, that can be used in commands:

• 0000:c0:11.0

• 0000:c0:11.1

• 0000:c0:11.2

• 0000:c0:11.3

• 0000:c0:11.4

• 0000:c0:11.5

e.g. -p_port 0000:c0:11.0.

TCP/UDP port

Note: In order to avoid port collisions, try to assign a separate and divergent set of ports for each container running on a server.

Use any unused set of ports out of the Well-known ports pool - 1024..65535, e.g. --expose=20000-20010.

Out of those specified ports, choose a single -udp_port, e.g. -udp_port 20000.

Final destination IP address

The -p_rx_ip parameter is used for the destination IP, e.g. -p_rx_ip 192.168.2.1.

GPU-related settings

This chapter explains how to gather the platform information regarding graphics card/s and use them in commands.

Gathering information about render device

Following commands can be used to determine the rendering device’s location:

devices_path="/dev/dri/"
readarray -t rendering_devices <<< $(ls -l "$devices_path" | awk '//{if($1 ~ "^c" && $4 == "render"){print $NF}}')
rendering_device=${devices_path}${rendering_devices[0]}
echo "Found: ${rendering_devices[@]}"
echo "Selected: ${rendering_device}"

Variable rendering_device holds the location value and can be used in further command building as a -qsv_device ${rendering_device}, e.g. -qsv_device /dev/dri/renderD128 (as a part of [broadcast_suite_parameters], -hwaccel qsv must be used as well in order to enable hardware acceleration).