|
A Discrete-Event Network Simulator
|
Models |
|
|
A Discrete-Event Network Simulator
|
Models |
This chapter describes the TCP models available in ns-3.
ns-3 was written to support multiple TCP implementations. The implementations inherit from a few common header classes in the src/network directory, so that user code can swap out implementations with minimal changes to the scripts.
There are two important abstract base classes:
There are presently three implementations of TCP available for ns-3.
It should also be mentioned that various ways of combining virtual machines with ns-3 makes available also some additional TCP implementations, but those are out of scope for this chapter.
Until ns-3.10 release, ns-3 contained a port of the TCP model from GTNetS. This implementation was substantially rewritten by Adriam Tam for ns-3.10. The model is a full TCP, in that it is bidirectional and attempts to model the connection setup and close logic.
The implementation of TCP is contained in the following files:
src/internet/model/tcp-header.{cc,h}
src/internet/model/tcp-l4-protocol.{cc,h}
src/internet/model/tcp-socket-factory-impl.{cc,h}
src/internet/model/tcp-socket-base.{cc,h}
src/internet/model/tcp-tx-buffer.{cc,h}
src/internet/model/tcp-rx-buffer.{cc,h}
src/internet/model/tcp-rfc793.{cc,h}
src/internet/model/tcp-tahoe.{cc,h}
src/internet/model/tcp-reno.{cc,h}
src/internet/model/tcp-westwood.{cc,h}
src/internet/model/tcp-newreno.{cc,h}
src/internet/model/rtt-estimator.{cc,h}
src/network/model/sequence-number.{cc,h}
Different variants of TCP congestion control are supported by subclassing the common base class TcpSocketBase. Several variants are supported, including RFC 793 (no congestion control), Tahoe, Reno, Westwood, Westwood+, and NewReno. NewReno is used by default. See the Usage section of this document for on how to change the default TCP variant used in simulation.
In many cases, usage of TCP is set at the application layer by telling the ns-3 application which kind of socket factory to use.
Using the helper functions defined in src/applications/helper and src/network/helper, here is how one would create a TCP receiver:
// Create a packet sink on the star "hub" to receive these packets
uint16_t port = 50000;
Address sinkLocalAddress(InetSocketAddress (Ipv4Address::GetAny (), port));
PacketSinkHelper sinkHelper ("ns3::TcpSocketFactory", sinkLocalAddress);
ApplicationContainer sinkApp = sinkHelper.Install (serverNode);
sinkApp.Start (Seconds (1.0));
sinkApp.Stop (Seconds (10.0));
Similarly, the below snippet configures OnOffApplication traffic source to use TCP:
// Create the OnOff applications to send TCP to the server
OnOffHelper clientHelper ("ns3::TcpSocketFactory", Address ());
The careful reader will note above that we have specified the TypeId of an abstract base class TcpSocketFactory. How does the script tell ns-3 that it wants the native ns-3 TCP vs. some other one? Well, when internet stacks are added to the node, the default TCP implementation that is aggregated to the node is the ns-3 TCP. This can be overridden as we show below when using Network Simulation Cradle. So, by default, when using the ns-3 helper API, the TCP that is aggregated to nodes with an Internet stack is the native ns-3 TCP.
To configure behavior of TCP, a number of parameters are exported through the ns-3 attribute system. These are documented in the Doxygen <http://www.nsnam.org/doxygen/classns3_1_1_tcp_socket.html> for class TcpSocket. For example, the maximum segment size is a settable attribute.
To set the default socket type before any internet stack-related objects are created, one may put the following statement at the top of the simulation program:
Config::SetDefault ("ns3::TcpL4Protocol::SocketType", StringValue ("ns3::TcpTahoe"));
For users who wish to have a pointer to the actual socket (so that socket operations like Bind(), setting socket options, etc. can be done on a per-socket basis), Tcp sockets can be created by using the Socket::CreateSocket() method. The TypeId passed to CreateSocket() must be of type ns3::SocketFactory, so configuring the underlying socket type must be done by twiddling the attribute associated with the underlying TcpL4Protocol object. The easiest way to get at this would be through the attribute configuration system. In the below example, the Node container “n0n1” is accessed to get the zeroth element, and a socket is created on this node:
// Create and bind the socket...
TypeId tid = TypeId::LookupByName ("ns3::TcpTahoe");
Config::Set ("/NodeList/*/$ns3::TcpL4Protocol/SocketType", TypeIdValue (tid));
Ptr<Socket> localSocket =
Socket::CreateSocket (n0n1.Get (0), TcpSocketFactory::GetTypeId ());
Above, the “*” wild card for node number is passed to the attribute configuration system, so that all future sockets on all nodes are set to Tahoe, not just on node ‘n0n1.Get (0)’. If one wants to limit it to just the specified node, one would have to do something like:
// Create and bind the socket...
TypeId tid = TypeId::LookupByName ("ns3::TcpTahoe");
std::stringstream nodeId;
nodeId << n0n1.Get (0)->GetId ();
std::string specificNode = "/NodeList/" + nodeId.str () + "/$ns3::TcpL4Protocol/SocketType";
Config::Set (specificNode, TypeIdValue (tid));
Ptr<Socket> localSocket =
Socket::CreateSocket (n0n1.Get (0), TcpSocketFactory::GetTypeId ());
Once a TCP socket is created, one will want to follow conventional socket logic and either connect() and send() (for a TCP client) or bind(), listen(), and accept() (for a TCP server). See Sockets APIs for a review of how sockets are used in ns-3.
Several TCP validation test results can be found in the wiki page describing this implementation.
The Network Simulation Cradle (NSC) is a framework for wrapping real-world network code into simulators, allowing simulation of real-world behavior at little extra cost. This work has been validated by comparing situations using a test network with the same situations in the simulator. To date, it has been shown that the NSC is able to produce extremely accurate results. NSC supports four real world stacks: FreeBSD, OpenBSD, lwIP and Linux. Emphasis has been placed on not changing any of the network stacks by hand. Not a single line of code has been changed in the network protocol implementations of any of the above four stacks. However, a custom C parser was built to programmatically change source code.
NSC has previously been ported to ns-2 and OMNeT++, and was was added to ns-3 in September 2008 (ns-3.2 release). This section describes the ns-3 port of NSC and how to use it.
To some extent, NSC has been superseded by the Linux kernel support within Direct Code Execution (DCE). However, NSC is still available through the bake build system. NSC supports Linux kernels 2.6.18 and 2.6.26, but newer versions of the kernel have not been ported.
Presently, NSC has been tested and shown to work on these platforms: Linux i386 and Linux x86-64. NSC does not support powerpc. Use on FreeBSD or OS X is unsupported (although it may be able to work).
Building NSC requires the packages flex and bison.
As of ns-3.17 or later, NSC must either be downloaded separately from its own repository, or downloading when using the bake build system of ns-3.
For ns-3.17 or later releases, when using bake, one must configure NSC as part of an “allinone” configuration, such as:
$ cd bake
$ python bake.py configure -e ns-allinone-3.19
$ python bake.py download
$ python bake.py build
Instead of a released version, one may use the ns-3 development version by specifying “ns-3-allinone” to the configure step above.
NSC may also be downloaded from its download site using Mercurial:
$ hg clone https://secure.wand.net.nz/mercurial/nsc
Prior to the ns-3.17 release, NSC was included in the allinone tarball and the released version did not need to be separately downloaded.
NSC may be built as part of the bake build process; alternatively, one may build NSC by itself using its build system; e.g.:
$ cd nsc-dev
$ python scons.py
Once NSC has been built either manually or through the bake system, change into the ns-3 source directory and try running the following configuration:
$ ./waf configure
If NSC has been previously built and found by waf, then you will see:
Network Simulation Cradle : enabled
If NSC has not been found, you will see:
Network Simulation Cradle : not enabled (NSC not found (see option --with-nsc))
In this case, you must pass the relative or absolute path to the NSC libraries with the “–with-nsc” configure option; e.g.
$ ./waf configure --with-nsc=/path/to/my/nsc/directory
For ns-3 releases prior to the ns-3.17 release, using the build.py script in ns-3-allinone directory, NSC will be built by default unless the platform does not support it. To explicitly disable it when building ns-3, type:
$ ./waf configure --enable-examples --enable-tests --disable-nsc
If waf detects NSC, then building ns-3 with NSC is performed the same way with waf as without it. Once ns-3 is built, try running the following test suite:
$ ./test.py -s ns3-tcp-interoperability
If NSC has been successfully built, the following test should show up in the results:
PASS TestSuite ns3-tcp-interoperability
This confirms that NSC is ready to use.
There are a few example files. Try:
$ ./waf --run tcp-nsc-zoo
$ ./waf --run tcp-nsc-lfn
These examples will deposit some .pcap files in your directory, which can be examined by tcpdump or wireshark.
Let’s look at the examples/tcp/tcp-nsc-zoo.cc file for some typical usage. How does it differ from using native ns-3 TCP? There is one main configuration line, when using NSC and the ns-3 helper API, that needs to be set:
InternetStackHelper internetStack;
internetStack.SetNscStack ("liblinux2.6.26.so");
// this switches nodes 0 and 1 to NSCs Linux 2.6.26 stack.
internetStack.Install (n.Get(0));
internetStack.Install (n.Get(1));
The key line is the SetNscStack. This tells the InternetStack helper to aggregate instances of NSC TCP instead of native ns-3 TCP to the remaining nodes. It is important that this function be called before calling the Install() function, as shown above.
Which stacks are available to use? Presently, the focus has been on Linux 2.6.18 and Linux 2.6.26 stacks for ns-3. To see which stacks were built, one can execute the following find command at the ns-3 top level directory:
$ find nsc -name "*.so" -type f
nsc/linux-2.6.18/liblinux2.6.18.so
nsc/linux-2.6.26/liblinux2.6.26.so
This tells us that we may either pass the library name liblinux2.6.18.so or liblinux2.6.26.so to the above configuration step.
NSC TCP shares the same configuration attributes that are common across TCP sockets, as described above and documented in Doxygen
Additionally, NSC TCP exports a lot of configuration variables into the ns-3 attributes system, via a sysctl-like interface. In the examples/tcp/tcp-nsc-zoo example, you can see the following configuration:
// this disables TCP SACK, wscale and timestamps on node 1 (the attributes
represent sysctl-values).
Config::Set ("/NodeList/1/$ns3::Ns3NscStack<linux2.6.26>/net.ipv4.tcp_sack",
StringValue ("0"));
Config::Set ("/NodeList/1/$ns3::Ns3NscStack<linux2.6.26>/net.ipv4.tcp_timestamps",
StringValue ("0"));
Config::Set ("/NodeList/1/$ns3::Ns3NscStack<linux2.6.26>/net.ipv4.tcp_window_scaling",
StringValue ("0"));
These additional configuration variables are not available to native ns-3 TCP.
Also note that default values for TCP attributes in ns-3 TCP may differ from the nsc TCP implementation. Specifically in ns-3:
- TCP default MSS is 536
- TCP Delayed Ack count is 2
Therefore when making comparisons between results obtained using nsc and ns-3 TCP, care must be taken to ensure these values are set appropriately. See /examples/tcp/tcp-nsc-comparision.cc for an example.
This subsection describes the API that NSC presents to ns-3 or any other simulator. NSC provides its API in the form of a number of classes that are defined in sim/sim_interface.h in the nsc directory.
The ns-3 implementation makes use of the above NSC API, and is implemented as follows.
The three main parts are:
src/internet/model/nsc-tcp-l4-protocol is the main class. Upon Initialization, it loads an nsc network stack to use (via dlopen()). Each instance of this class may use a different stack. The stack (=shared library) to use is set using the SetNscLibrary() method (at this time its called indirectly via the internet stack helper). The nsc stack is then set up accordingly (timers etc). The NscTcpL4Protocol::Receive() function hands the packet it receives (must be a complete tcp/ip packet) to the nsc stack for further processing. To be able to send packets, this class implements the nsc send_callback method. This method is called by nsc whenever the nsc stack wishes to send a packet out to the network. Its arguments are a raw buffer, containing a complete TCP/IP packet, and a length value. This method therefore has to convert the raw data to a Ptr<Packet> usable by ns-3. In order to avoid various ipv4 header issues, the nsc ip header is not included. Instead, the tcp header and the actual payload are put into the Ptr<Packet>, after this the Packet is passed down to layer 3 for sending the packet out (no further special treatment is needed in the send code path).
This class calls ns3::NscTcpSocketImpl both from the nsc wakeup() callback and from the Receive path (to ensure that possibly queued data is scheduled for sending).
src/internet/model/nsc-tcp-socket-impl implements the nsc socket interface. Each instance has its own nscTcpSocket. Data that is Send() will be handed to the nsc stack via m_nscTcpSocket->send_data(). (and not to nsc-tcp-l4, this is the major difference compared to ns-3 TCP). The class also queues up data that is Send() before the underlying descriptor has entered an ESTABLISHED state. This class is called from the nsc-tcp-l4 class, when the nsc-tcp-l4 wakeup() callback is invoked by nsc. nsc-tcp-socket-impl then checks the current connection state (SYN_SENT, ESTABLISHED, LISTEN...) and schedules appropriate callbacks as needed, e.g. a LISTEN socket will schedule Accept to see if a new connection must be accepted, an ESTABLISHED socket schedules any pending data for writing, schedule a read callback, etc.
Note that ns3::NscTcpSocketImpl does not interact with nsc-tcp directly: instead, data is redirected to nsc. nsc-tcp calls the nsc-tcp-sockets of a node when its wakeup callback is invoked by nsc.
For more information, see this wiki page.