View source for UDPTX ← UDPTX

You do not have permission to edit this page, for the following reason:

The action you have requested is limited to users in the group: Users.

You can view and copy the source of this page.

Return to UDPTX.

Retrieved from "https://wiki.brandmeister.network/index.php/UDPTX"

UDPTX is BrandMeister-own UDP communication library, used to transmit and receive UDP traffic fast. It is very important for BrandMeister to spend less time to send and receive packets, it makes transmission (and finally sound) more smooth.

UDP Transmitter[edit]

At this moment BrandMeister provides several backends (options) to send outgoing UDP:


UDPTX.png

socket[edit]

This is standard default backend that uses Berkley sockets for sending a traffic. It tries to send the data in non-blocking mode and has special transmission thread to re-send failed packets or offload main thread on high load (> 50% CPU core).

You have to use it if you have[edit]

Performance[edit]

Passed performance tests to transmit to 20K connections

Configuration[edit]

transmitter = "socket";

uring[edit]

This is standard default backend for versions builT for Debian 12 and higher. Uses Berkley sockets and URing API for sending a traffic.

Configuration[edit]

transmitter = "uring";

raw[edit]

This is fast forwarding backend that uses RAW (PACKET_MMAP) socket of Ethernet interface for sending a traffic. It allows to save up to 50% CPU time and has great compatibility.

Limitations[edit]

Performance[edit]

Passed performance tests to transmit to 40K connections

Configuration[edit]

transmitter = "raw:<interface name> <module parameters>";

transmitter = "raw:eth0 --count 2";

Module parameters[edit]

All these parameters are optional and override default settings 

(-b) --bypass       - bypass QDisc traffic control (PACKET_QDISC_BYPASS)
(-c) --count   <n>  - count of transmitter cores (default is 1)

Note: in case when you need to use VLANs, you have pass a physical interface (eth0 for example) as interface to bind and use --reference-interface key to pass the name of VLAN interface (eth0.1 for example)

AF_XDP[edit]

This is faster forwarding backend that uses AF_XDP socket of Ethernet interface for sending a traffic and in most cases communicates directly with Linux network interface driver. Due to monopolise NIC queues any configuration with multiple cores has to use ranges of queues per instance (for example: instance 1 uses 8 queues staring from 0 and instance 2 uses 8 queues starting from 8).

Limitations[edit]

Performance[edit]

Passed performance tests to transmit to 55K connections

Configuration[edit]

transmitter = "xdp:<interface name> <module parameters>";

transmitter = "xdp:eth0";

Module parameters[edit]

All these parameters are optional and override default settings 

(-w) --wakeup                   - use XDP in wakeup mode (XDP_USE_NEED_WAKEUP)
(-i) --in-flight           <n>  - set number of packets in flight (default value is calculated)
(-c) --core-ratio          <n>  - ratio between NIC queues and transmitting cores (instead of default value of 2)
(-f) --first-queue         <n>  - set number of first transmitting queue manually (default is 0)
(-q) --queue-count         <n>  - set transmission queue count manually (instead of all available queues)
(-l) --buffer-length       <n>  - set workers buffer length to <n> slots (instead of default value of 1024)
(-r) --reference-interface <s>  - reference interface to monitor routes, gateways

Note: in case when you need to use VLANs, you have pass a physical interface (eth0 for example) as interface to bind and use --reference-interface key to pass the name of VLAN interface (eth0.1 for example)

DPDK[edit]

This is fastest forwarding backend that uses kernel-bypass NIC driver for sending a traffic. It allows to save much more CPU time due to direct poll communications to the NIC and CRC offload features of some NIC models. In some tests we got up to 75% acceleration. List of supported NIC models can be found here.

Limitations[edit]

Performance[edit]

Passed performance tests to transmit to 100K connections on ixgbe

Configuration[edit]

transmitter = "Modules/DPDK-edge.so:<reference interface> <EAL parameters> [--] <module parameters>";

transmitter = "Modules/DPDK-edge.so:eth0 -a 0000:af:00.0 --file-prefix bm --lcores '(0-8)@1,3,5,7,9' -- -c 1 -q 2048 -b 1 -l 4096"; 

# /etc/systemd/system/brandmeister@.service.d/override.conf
[Service]
User=root

Module parameters[edit]

All these parameters are optional and override default settings 

(-c) --core-ratio    <n>  - ratio between NIC queues and DPDK cores (instead of default value of 4)
(-q) --queue-size    <n>  - set PMD queue size to <n> slots (instead of automatically generated)
(-l) --buffer-length <n>  - set workers buffer length to <n> slots (instead of default value of 512)
(-p) --pthresh     <n> |
(-h) --hthresh     <n> | PMD specific threshold values:
(-w) --wthresh     <n> | https://doc.dpdk.org/guides/prog_guide/poll_mode_drv.html#configuration-of-transmit-queues
(-r) --rs-thresh   <n> |
(-f) --free-thresh <n> |
(-s) --software-crc       - force software CRC calculation

Ixy[edit]

Ixy is very experimental and light user-space network driver. At this moment it supports Intel 82599ES family (aka Intel X520) and virtio. Please read Ixy documentation.

Limitations[edit]

Performance[edit]

Passed performance tests to transmit 120K connections

Configuration[edit]

transmitter = "Modules/Dixie.so:<reference interface> <PCI address> <module parameters>"; 

--core-ratio    <n>
--queue-size    <n>
--queue-count   <n>
--buffer-length <n>

transmitter = "Modules/Dixie.so:eth0 0000:af:00.0 --queue-count 8";'

verbs[edit]

This is method is only suitable when you use OFED/RDMA-enabled NIC, mostly Mellanox and its OEM derives. Both IP over Ethernet and IP over InfiniBand are supported.

Configuration[edit]

transmitter = "verbs:<interface name> <module parameters>";

transmitter = "verbs:rocep175s0 --reference-interface vlan100";

Module parameters[edit]

All these parameters are optional and override default settings 

--core-count          <n>       - set count of workers (default is 1)
--queue-size          <n>       - set queue size manually (default is 512)
--device-port         <n>       - device port (default is 1)
--buffer-length       <n>       - set workers buffer length to <n> slots (default is 1024)
--reference-interface <ipoib0>  - reference interface to monitor routes, gateways

UDP Receiver[edit]

In reception part UDPTX's driver works in parallel with socket receiver. All it does, is accelerate reception of UDP packets on particular interface.

socket[edit]

This method allows Core to offload main thread on high-loaded servers by reducing amount of system calls to receive UDP messages.

Configuration[edit]

receiver = "socket <module parameters>";

Module parameters[edit]

All these parameters are optional and override default settings 

(-l) --buffer-length       <n>  - set workers buffer length to <n> slots (instead of default value of 1024)

uring[edit]

This method allows Core to offload main thread on high-loaded servers by reducing amount of system calls to receive UDP messages. Available in builds for Debian 12 and higher.

Configuration[edit]

receiver = "uring <module parameters>";

Module parameters[edit]

All these parameters are optional and override default settings 

(-l) --buffer-length       <n>  - set buffer length to <n> slots (instead of default value of <fds limit>)

eBPF + AF_XDP[edit]

This is modern method to accelerate UDP reception in BrandMeister Core. It allows to save up to 30% CPU time.

Limitations[edit]

Configuration[edit]

receiver = "Modules/ExpressFilter.o:<interface name> <module parameters>";

receiver = "Modules/ExpressFilter.o:eth0 --reference-interface eth0.1";

Module parameters[edit]

All these parameters are optional and override default settings 

(-w) --wakeup                   - use XDP in wakeup mode (XDP_USE_NEED_WAKEUP)
(-f) --first-queue         <n>  - set number of first receiving queue manually (default is 0)
(-q) --queue-count         <n>  - set receiving queue count manually (instead of all available queues)
(-l) --buffer-length       <n>  - set workers buffer length to <n> slots (instead of default value of 1024)
(-r) --reference-interface <s>  - reference interface to monitor routes, gateways

Note: in case when you need to use VLANs, you have pass a physical interface (eth0 for example) as interface to bind and use --reference-interface key to pass the name of VLAN interface (eth0.1 for example)

eBPF + AF_XDP + XDPHelper[edit]

This is method is fully the same as eBPF + AF_XDP but uses small additional daemon XDPHelper to load and share eBPF program between several BrandMeister Core instances. This method also helps when NIC resets on eBPF load such as isgbe. XDPHelper is supplied with BrandMeister Core and starts automatically only when required (thanks to systemd and D-BUS activation). By default XDPHelper uses eBPF program ExpressFilter.o (see xdphelper.service).

Limitations[edit]

Configuration[edit]

receiver = "xdp:<interface name> <module parameters>";

receiver = "xdp:eth0 --reference-interface eth0.1";

Module parameters[edit]

All these parameters are optional and override default settings 

(-w) --wakeup                   - use XDP in wakeup mode (XDP_USE_NEED_WAKEUP)
(-f) --first-queue         <n>  - set number of first receiving queue manually (default is 0)
(-q) --queue-count         <n>  - set receiving queue count manually (instead of all available queues)
(-l) --buffer-length       <n>  - set workers buffer length to <n> slots (instead of default value of 1024)
(-r) --reference-interface <s>  - reference interface to monitor routes, gateways

Note: in case when you need to use VLANs, you have pass a physical interface (eth0 for example) as interface to bind and use --reference-interface key to pass the name of VLAN interface (eth0.1 for example)

verbs[edit]

This is method is only suitable when you use OFED/RDMA-enabled NIC, mostly Mellanox and its OEM derives. NIC should have support of Flow Steering. Both IP over Ethernet and IP over InfiniBand are supported. In case of Ethernet card it should have N-tuples enabled (ethtool -K eth0 ntuple on). Mellanox Connectx cards need device managed flow steering (options mlx4_core log_num_mgm_entry_size=-1)

Configuration[edit]

receiver = "verbs:<interface name> <module parameters>";

receiver = "verbs:rocep175s0 --reference-interface vlan100";

Module parameters[edit]

All these parameters are optional and override default settings 

--queue-size          <n>       - set queue size manually (default is 512)
--reference-interface <eth0.1>  - reference interface to monitor routes, gateways