nix-vector-routing: Update doc on IPv6, examples and usage
This commit is contained in:
Ameya Deshpande
committed by
Tommaso Pecorella
Tommaso Pecorella
parent
a80d4c95e0
commit
ec05adaa9f
BIN
src/nix-vector-routing/doc/figures/nms.png
Normal file
BIN
src/nix-vector-routing/doc/figures/nms.png
Normal file
Binary file not shown.
|
After Width: | Height: | Size: 52 KiB |
@@ -28,27 +28,52 @@ a breadth-first search and an efficient route-storage data structure
|
||||
known as a nix-vector.
|
||||
|
||||
When a packet is generated at a node for transmission, the route is
|
||||
calculated, and the nix-vector is built.
|
||||
The nix-vector stores an index for each hop along the path, which
|
||||
corresponds to the neighbor-index. This index is used to determine
|
||||
which net-device and gateway should be used. To route a packet, the
|
||||
nix-vector must be transmitted with the packet. At each hop, the
|
||||
current node extracts the appropriate neighbor-index from the
|
||||
nix-vector and transmits the packet through the corresponding
|
||||
net-device. This continues until the packet reaches the destination.
|
||||
calculated, and the nix-vector is built.
|
||||
|
||||
NixVectorRouting bases its routing decisions on the nodes addresses,
|
||||
and it does **not** check that the nodes are in the proper subnets.
|
||||
In other terms, using NixVectorRouting you could have an (apparently)
|
||||
working network that violates every good practice in IP address assignments.
|
||||
**How is the Nix-Vector calculated?**
|
||||
The nix-vector stores an index for each hop along the path, which
|
||||
corresponds to the neighbor-index. This index is used to determine
|
||||
which net-device and gateway should be used.
|
||||
|
||||
**How does the routing take place?**
|
||||
To route a packet, the nix-vector must be transmitted with the packet.
|
||||
At each hop, the current node extracts the appropriate neighbor-index
|
||||
from the nix-vector and transmits the packet through the corresponding
|
||||
net-device. This continues until the packet reaches the destination.
|
||||
|
||||
.. note::
|
||||
Nix-Vector routing does not use any routing metrics (interface metrics)
|
||||
during the calculation of nix-vector. It is only based on the shortest
|
||||
path calculated according to BFS.
|
||||
|
||||
**How does Nix decide between two equally short path from source to
|
||||
destination?**
|
||||
It depends on how the topology is constructed i.e., the order in which the
|
||||
net-devices are added on a node and net-devices added on the channels
|
||||
associated with current node's net-devices. Please check the ``nix-simple.cc``
|
||||
example below to understand how nix-vectors are calculated.
|
||||
|
||||
|
||||
|ns3| supports IPv4 as well as IPv6 Nix-Vector routing.
|
||||
|
||||
Scope and Limitations
|
||||
=====================
|
||||
|
||||
Currently, the ns-3 model of nix-vector routing supports IPv4 p2p links
|
||||
as well as CSMA links. It does not (yet) provide support for
|
||||
efficient adaptation to link failures. It simply flushes all nix-vector
|
||||
routing caches. Finally, IPv6 is not supported.
|
||||
Currently, the |ns3| model of nix-vector routing supports IPv4 and IPv6
|
||||
p2p links as well as CSMA links. It does not (yet) provide support for
|
||||
efficient adaptation to link failures. It simply flushes all nix-vector
|
||||
routing caches.
|
||||
|
||||
NixVectorRouting bases its routing decisions on the nodes' interface
|
||||
addresses and does **not** check whether the nodes are in the proper
|
||||
subnets or the addresses have been appropriately assigned. In other
|
||||
terms, using Nix-Vector routing, it is possible to have a working
|
||||
network that violates every good practice in IP address assignments.
|
||||
|
||||
In case of IPv6, Nix assumes the link-local addresses assigned are **unique**.
|
||||
When using the IPv6 stack, the link-local address allocation is unique by
|
||||
default over the entire topology. However, if the link-local addresses are
|
||||
assigned manually, the user must ensure uniqueness of link-local addresses.
|
||||
|
||||
|
||||
Usage
|
||||
@@ -57,8 +82,40 @@ Usage
|
||||
The usage pattern is the one of all the Internet routing protocols.
|
||||
Since NixVectorRouting is not installed by default in the
|
||||
Internet stack, it is necessary to set it in the Internet Stack
|
||||
helper by using ``InternetStackHelper::SetRoutingHelper``
|
||||
helper by using ``InternetStackHelper::SetRoutingHelper``.
|
||||
|
||||
Remember to include the header file ``ns3/nix-vector-helper.h`` to
|
||||
use IPv4 or IPv6 Nix-Vector routing.
|
||||
|
||||
.. note::
|
||||
The previous header file ``ns3/ipv4-nix-vector-helper.h`` is
|
||||
maintained for backward compatibility reasons. Therefore, the
|
||||
existing IPv4 Nix-Vector routing simulations should work fine.
|
||||
|
||||
* Using IPv4 Nix-Vector Routing:
|
||||
|
||||
.. code-block:: c++
|
||||
|
||||
Ipv4NixVectorHelper nixRouting;
|
||||
InternetStackHelper stack;
|
||||
stack.SetRoutingHelper (nixRouting); // has effect on the next Install ()
|
||||
stack.Install (allNodes); // allNodes is the NodeContainer
|
||||
|
||||
* Using IPv6 Nix-Vector Routing:
|
||||
|
||||
.. code-block:: c++
|
||||
|
||||
Ipv6NixVectorHelper nixRouting;
|
||||
InternetStackHelper stack;
|
||||
stack.SetRoutingHelper (nixRouting); // has effect on the next Install ()
|
||||
stack.Install (allNodes); // allNodes is the NodeContainer
|
||||
|
||||
.. note::
|
||||
The NixVectorHelper helper class helps to use NixVectorRouting functionality.
|
||||
The NixVectorRouting model can also be used directly to use Nix-Vector routing.
|
||||
To use it directly, the header file ``ns3/nix-vector-routing.h`` should be
|
||||
included. The previous header file ``ns3/ipv4-nix-vector-routing.h`` is maintained
|
||||
for backwards compatibility with any existing IPv4 Nix-Vector routing simulations.
|
||||
|
||||
Examples
|
||||
========
|
||||
@@ -66,3 +123,90 @@ Examples
|
||||
The examples for the NixVectorRouting module lives in
|
||||
the directory ``src/nix-vector-routing/examples``.
|
||||
|
||||
There are examples which use both IPv4 and IPv6 networking.
|
||||
|
||||
* nix-simple.cc
|
||||
|
||||
::
|
||||
|
||||
/*
|
||||
* ________
|
||||
* / \
|
||||
* n0 -- n1 -- n2 -- n3
|
||||
*
|
||||
* n0 IP: 10.1.1.1, 10.1.4.1
|
||||
* n1 IP: 10.1.1.2, 10.1.2.1
|
||||
* n2 IP: 10.1.2.2, 10.1.3.1, 10.1.4.2
|
||||
* n3 IP: 10.1.3.2
|
||||
*/
|
||||
|
||||
In this topology, we install Nix-Vector routing between source
|
||||
n0 and destination n3. The shortest possible route will be
|
||||
n0 -> n2 -> n3.
|
||||
|
||||
Let's see how the nix-vector will be generated for this path:
|
||||
|
||||
n0 has 2 neighbors i.e. n1 and n3. n0 is connected to both using
|
||||
separate net-devices. But the net-device for n0 -- n1 p2p link was
|
||||
created before the netdevice for n0 -- n2 p2p link. Thus, n2 has
|
||||
neighbor-index of 1 (n1 has 0) with respect to n0.
|
||||
|
||||
n2 has 3 neighbors i.e. n1, n3 and n0. The n2 net-device for n1 -- n2
|
||||
p2p link was created before the n2 net-device for n2 -- n3 p2p link
|
||||
which was before the n2 netdevice for n0 -- n2 p2p link. This, n3
|
||||
has neighbor-index of 01 (n1 has 00 and n0 has 10) with repect to n2.
|
||||
|
||||
Thus, the nix-vector for the path from n0 to n3 is 101.
|
||||
|
||||
.. note::
|
||||
This neighbor-index or nix-index has total number of bits equal to
|
||||
minimum number of bits required to represent all the neighbors in
|
||||
their binary form.
|
||||
|
||||
.. note::
|
||||
If there are multiple netdevices connected to the current netdevice
|
||||
on the channel then it depends on which order netdevices were added
|
||||
to the channel.
|
||||
|
||||
#. Using IPv4:
|
||||
.. code-block:: bash
|
||||
|
||||
# By default IPv4 network is selected
|
||||
./waf --run src/nix-vector-routing/examples/nix-simple
|
||||
|
||||
#. Using IPv6:
|
||||
.. code-block:: bash
|
||||
|
||||
# Set ip as "v6" explicity
|
||||
./waf --run "src/nix-vector-routing/examples/nix-simple --useIPv6"
|
||||
|
||||
* nms-p2p-nix.cc
|
||||
|
||||
This example demonstrates the advantage of Nix-Vector routing as Nix
|
||||
performs source-based routing (BFS) to have faster routing.
|
||||
|
||||
.. image:: figures/nms.png
|
||||
:alt: NMS P2P Network Diagram
|
||||
|
||||
#. Using IPv4:
|
||||
.. code-block:: bash
|
||||
|
||||
# By default IPv4 network is selected
|
||||
./waf --run src/nix-vector-routing/examples/nms-p2p-nix
|
||||
|
||||
#. Using IPv6:
|
||||
.. code-block:: bash
|
||||
|
||||
# Set ip as "v6" explicity
|
||||
./waf --run "src/nix-vector-routing/examples/nms-p2p-nix --useIPv6"
|
||||
|
||||
* nix-simple-multi-address.cc
|
||||
|
||||
This is an IPv4 example demonstrating multiple interface addresses. This
|
||||
example also shows how address assignment in between the simulation causes
|
||||
the all the route caches and Nix caches to flush.
|
||||
|
||||
.. code-block:: bash
|
||||
|
||||
# By default IPv4 network is selected
|
||||
./waf --run src/nix-vector-routing/examples/nix-simple-multi-address
|
||||
|
||||
Reference in New Issue
Block a user