mirror of
https://github.com/rtbrick/bngblaster.git
synced 2024-05-06 15:54:57 +00:00
Christian Giese
52 lines
3.4 KiB
Plaintext
52 lines
3.4 KiB
Plaintext
Architecture
|
||
------------
|
||
|
||
The BNG Blaster has been completely built from scratch in **C**. This includes user-space implementations
|
||
of the entire protocol stack. Its core is based on a very simple event loop that serves timers and
|
||
signals. The timers have been built using a lightweight constant time (*O(1)*) library. The timer library
|
||
was built to start, restart and delete the protocol session FSM timers quickly and at scale.
|
||
|
||
The BNG Blaster expects a Linux kernel network interface that is up but not configured with any IP addresses
|
||
or VLAN as it expects to receive and transmit RAW ethernet packets.
|
||
|
||
The BNG Blaster does I/O using high-speed polling timers with a mix of Linux
|
||
`RAW Packet Sockets <https://man7.org/linux/man-pages/man7/packet.7.html>`_ and
|
||
`Packet MMAP <https://www.kernel.org/doc/html/latest/networking/packet_mmap.html>`_.
|
||
|
||
The second one is a so-called PACKET_RX_RING/PACKET_TX_RING abstraction where a user-space program gets a fast
|
||
lane into reading and writing to kernel interfaces using a shared ring buffer. The shared ring buffer is a
|
||
memory-mapped window shared between the kernel and the user space. This low overhead abstraction allows us to
|
||
transmit and receive traffic without doing expensive system calls. Sending and transmitting traffic via Packet MMAP is
|
||
as easy as copying a packet into a buffer and setting a flag.
|
||
|
||
.. image:: ../images/bbl_arch.png
|
||
:alt: BNG Blaster Architecture
|
||
|
||
The BNG Blaster supports many configurable I/O modes listed with ``bngblaster -v`` but except for the default
|
||
mode ``packet_mmap_raw`` all other modes are currently considered experimental. In the default mode, all
|
||
packets are received in a Packet MMAP ring buffer and sent through RAW packet sockets. This combination
|
||
was the most efficient in our benchmark tests.
|
||
|
||
BNG Blaster's primary design goal is to simulate thousands of subscriber CPE’s with a small hardware resource
|
||
footprint. Simple to use and easy to integrate into our robot test automation infrastructure. This allows for
|
||
the simulation of massive PPPoE or IPoE (DHCP) subscribers including IPTV, traffic verification, and convergence
|
||
testing from a single medium-scale virtual machine or directly from a laptop.
|
||
|
||
The BNG Blaster provides three types of interface functions. The first interface function is called the access which
|
||
emulates the PPPoE or IPoE sessions. The second interface function is called network. This is used for
|
||
emulating the core-facing side of the internet with optional routing protocols. The last type is called a10nsp
|
||
interface which emulates a layer two provider interface. The term A10 refers to the end-to-end ADSL network
|
||
reference model from TR-025.
|
||
|
||
.. image:: ../images/bbl_interfaces.png
|
||
:alt: BNG Blaster Interfaces
|
||
|
||
This allows for verification of IP reachability by sending bidirectional traffic between all sessions
|
||
on the access interface and the network interface. The network interface is also used to inject downstream
|
||
multicast test traffic for IPTV tests. It is also possible to send RAW traffic streams between network
|
||
interfaces without any access interface defined for non-BNG testing.
|
||
|
||
One popular example of non-BNG tests with the BNG Blaster is the verification of a BGP full table by injecting
|
||
around 1M prefixes and setting up traffic streams for all prefixes with at least 1 PPS (1M PPS).
|
||
The BNG Blaster can verify and analyze every single flow with detailed per-flow statistics
|
||
(receive rate, loss, latency, …). |