Gst-nvdsudpsink
The Gst-nvdsudpsink plugin is a sink type component which can be used to transmit the RTP packets over IP network. Internally, the plugin uses Rivermax SDK APIs for network communications. NVIDIA Rivermax® offers a unique IP-based solution for any media and data streaming use case. For more details see the Rivermax Product Page.
nvdsudpsink
can operate in three different modes based on the input type and properties set to the component.
- RTP packets with Rivermax Generic API
This is default mode. In this mode nvdsudpsink will receive RTP packets and transmit it to network using Rivermax generic APIs. nvdsudpsink doesn’t require any information about the content of RTP packets.
- RTP packets with Rivermax Media API
This mode is for specific use case of SMPTE 2110-20, 30 streams. In this mode, nvdsudpsink supports SMPTE 2110-20 and SMPTE 2110-30 streams with traffic shaping. nvdsudpsink expects RTP packets having either uncompressed audio or video as payload and It will transmit those packets to network using Rivermax media APIs.
Since Rivermax media APIs need to know the information about the data being transmitted for it to handle the traffic shaping, nvdsudpsink needs that information for Rivermax initialization in the form of SDP file.
This mode can be enabled by setting
sdp-file
property of the component. In this mode, for uncompressed video streams,packets-per-line
andpayload-size
properties along withsdp-file
must be set. Similarly, value ofmtu
property of payloader component must be same aspayload-size
ofnvdsudpsink
.For uncompressed audio streams in this mode,
max-ptime
andptime-multiple
properties of payloader component must be set.
- Uncompressed data with Rivermax media API
In this mode, nvdsudpsink expects uncompressed audio or video frames directly instead of RTP packets. The component will convert those frames to RTP packets and then transmit over network using Rivermax media API.
This mode will be enabled if in addition to the requirements of mode #2 the incoming data is uncompressed video or audio frames instead of RTP packets.
In this mode, if
render-thread-core
property is set, a separate thread will be created to packetize the frame and transmit the packets. That thread will also be bound to the CPU core(s) provided as the value of the property.
Note
In this mode as uncompressed video input, only YUV 4:2:2 and RGB with 8 or 10 bit depth is supported.
Payload size and packets per line for some of common formats
To calculate the payload size and packets per line, use the following steps:
Calculate line stride as follows:
line stride = (width * pgroup size) / pgroup coverage
Divide this stride into number of packets having equal bytes. This will be value of packet-per-line.
Add 20 bytes to size of each packet calculated in step 2 for RTP + Payload header. This will be value for payload-size.
Note
It is assumed that there is only single Sample Row Data per RTP packet and line stride is equally divisible.
Following are values for some of the common formats:
YCbCr-4:2:2 1920x1080 8 bit: payload-size:1300, packets-per-line:3 YCbCr-4:2:2 1920x1080 10 bit: payload-size:1220, packets-per-line:4 YCbCr-4:2:0 1920x1080 8 bit: payload-size:1460, packets-per-line:2 YCbCr-4:2:0 1920x1080 10 bit: payload-size:1220, packets-per-line:3 RGB 1920x1080 8 bit: payload-size:1460, packets-per-line:4 RGB 1920x1080 10 bit: payload-size:1220, packets-per-line:6
Refer to the SMPTE ST 2110-20:2017
for more information.
System can also have Gstreamer provided OSS implementation of udp sink (udpsink
) component. In that case system would have two implementations for udp sink - udpsink
and nvdsudpsink
.
nvdsudpsink
component can only be used with NVIDIA ConnectX-5 and above cards after having installed Rivermax
SDK and its license.
Download and setup the
Rivermax
1.40.x SDK here: https://developer.nvidia.com/networking/rivermax-getting-started.Follow the instruction on the SDK page to obtain development license for Rivermax.
To select
nvdsudpsink
out of two installations, use eitherLOCAL_IFACE_IP
environment variable orlocal-iface-ip
property. Use the command below to export the environment variable:export LOCAL_IFACE_IP=<IP of NIC>
nvdsudpsink
component also requiresCAP_NET_RAW
capability. Either run the application that usesnvdsudpsink
component with superuser privilege or set theCAP_NET_RAW
capabilities using the following command.sudo setcap CAP_NET_RAW=ep <absolute path of application>
For example:
sudo setcap CAP_NET_RAW=ep /opt/nvidia/deepstream/deepstream/bin/deepstream-app sudo setcap CAP_NET_RAW=ep /usr/bin/gst-launch-1.0
Inputs and Outputs
Inputs
GstBufferList having RTP packets as buffer content.
GstBuffer having uncompressed audio or video frame.
Control parameters
LOCAL_IFACE_IP ENV flag or local-iface-ip property
host
port
chunk-size
packets-per-chunk
payload-size
sdp-file
packets-per-line
internal-thread-core
ptp-src
render-thread-core
Output
None
Features
The following table summarizes the features of the plugin.
Feature |
Description |
Release |
---|---|---|
Supports Rivermax Generic and Media APIs |
By default, it supports generic APIs, but media API can be enabled by setting sdp-file property. |
DS 6.1 |
Supports any type of RTP packet (Compressed, Uncompressed, audio etc.) with Generic APIs |
No restriction on content of RTP payload in case of Generic APIs. |
DS 6.1 |
Supports RTCP packets,In addition to RTP |
RTCP packets can also be transmitted. |
DS 6.1 |
Traffic shaping in case of Uncompressed streams (video and audio) |
By using Rivermax media APIs,packet pacing can be enabled for uncompressed video and audio streams. |
DS 6.1 |
Supports SMPTE 2110-20 |
Uncompressed SMPTE 2110-20 streams with packet pacing is supported through Rivermax media APIs. |
DS 6.1 |
Supports SMPTE 2110-30 |
Uncompressed SMPTE 2110-30 stream with packet pacing is supported through Rivermax media APIs. |
DS 6.1 |
Supports RTP packetization of uncompressed video |
Uncompressed video as input can be converted to RTP packets as per SMPTE 2110-20 before sending to network |
DS 6.2 |
Supports RTP packetization of uncompressed audio |
Uncompressed audio as input can be converted to RTP packets as per SMPTE 2110-30 before sending to network |
DS 6.2 |
Supports PTP clock |
In addition to system clock, hardware clock from the PTP source (Network card) can be used to generate RTP packet timestamp |
DS 6.2 |
Supports render thread affinity |
Separate thread to packetize and transmit the data. This thread can be bound to specific CPU core(s) |
DS 6.2 |
Gst Properties
The following table describes the Gst-nvdsudpsink plugin’s Gst properties.
Property |
Meaning |
Type and Range |
Example / Notes |
Platforms |
---|---|---|---|---|
port |
The port number to send the RTP packets to |
Integer, 0 to 65535 |
Port=5004 |
dGPU Jetson |
host |
Unicast / Multicast address to send the packets |
String |
address=192.168.4.60 |
dGPU Jetson |
chunk-size |
Number of chunks to allocate in case of generic APIs |
Integer, 1 to 65535 |
chunk-size=100 |
dGPU Jetson |
packets-per-chunk |
Number of packets per chunk |
Integer, 1 to 65535 |
packets-per-chunk=10 |
dGPU Jetson |
payload-size |
Size of payload in RTP packet |
Integer, 0 to 65535 |
payload-size=1400 |
dGPU Jetson |
packets-per-line |
Number of packets per line in case of media APIs |
Integer, 1 to 65535 |
packets-per-line=4 |
dGPU Jetson |
sdp-file |
Path to SDP file, to enable the use of Rivermax media APIs |
String |
sdp-file=”sdp_video.txt” |
dGPU Jetson |
local-iface-ip |
IP Address associated with network interface through which to transmit the data |
String |
local-iface-ip=”192.168.2.20” |
dGPU Jetson |
internal-thread-core |
CPU core to run Rivermax internal thread |
Integer, -1 to 1023 |
internal-thread-core=2 |
dGPU Jetson |
ptp-src |
IP Address of PTP source |
String |
ptp-src=”192.168.2.20” |
dGPU Jetson |
render-thread-core |
Comma seperated list of CPU core(s) for rendering thread |
String |
render-thread-core=”2,3” |
dGPU Jetson |
Example pipelines
Pipeline to transmit 24 bit 2 channel 48k audio stream using Rivermax media API:
gst-launch-1.0 audiotestsrc wave=8 ! 'audio/x-raw, format=(string)S24BE, rate=48000, channels=2, payload=(int)97' ! rtpL24pay ptime-multiple=1000000 max-ptime=1000000 seqnum-offset=0 pt=97 ! nvdsudpsink host=<ip address> port=<port number> local-iface-ip=<ip addr of NIC> sdp-file=<sdp file> internal-thread-core=4 -v --gst-debug=3
Pipeline with separate payload component to transmit 10bit YUV 4:2:2 1080p30 video stream using Rivermax media API:
gst-launch-1.0 videotestsrc ! 'video/x-raw, width=1920, height=1080, format=(string)UYVP, framerate=30/1' ! queue ! rtpvrawpay mtu=1220 ! nvdsudpsink host=<ip address> port=<port number> local-iface-ip=<ip addr of NIC> payload-size=1220 packets-per-line=4 sdp-file=<sdp file> -v --gst-debug=3
Pipeline without additional payload component to transmit 10bit YUV 4:2:2 1080p60 video stream using Rivermax media API:
gst-launch-1.0 videotestsrc ! 'video/x-raw, width=1920, height=1080, format=(string)UYVP, framerate=60/1' ! queue ! nvdsudpsink host=<ip address> port=<port number> local-iface-ip=<ip addr of NIC> payload-size=1220 packets-per-line=4 sdp-file=<sdp file> -v --gst-debug=3
Pipeline that uses PTP hardware clock from NIC:
gst-launch-1.0 videotestsrc ! 'video/x-raw, width=1920, height=1080, format=(string)UYVP, framerate=60/1' ! queue ! nvdsudpsink host=<ip address> port=<port number> local-iface-ip=<ip addr of NIC> payload-size=1220 packets-per-line=4 sdp-file=<sdp file> ptp-src=<ip addr of PTP source> -v --gst-debug=3
Pipeline that uses separate render thread and bind it to CPU core(s):
gst-launch-1.0 videotestsrc ! 'video/x-raw, width=1920, height=1080, format=(string)UYVP, framerate=60/1' ! queue ! nvdsudpsink host=<ip address> port=<port number> local-iface-ip=<ip addr of NIC> payload-size=1220 packets-per-line=4 sdp-file=<sdp file> ptp-src=<ip addr of PTP source> render-thread-core="list of core(s)" sync=0 -v --gst-debug=3