6.2 Models, Attributes and Reality

This is a convenient place to make a small excursion and make an important point. It may or may not be obvious to you, but whenever one is using a simulation, it is important to understand exactly what is being modeled and what is not. It is tempting, for example, to think of the CSMA devices and channels used in the previous section as if they were real Ethernet devices; and to expect a simulation result to directly reflect what will happen in a real Ethernet. This is not the case.

A model is, by definition, an abstraction of reality. It is ultimately the responsibility of the simulation script author to determine the so-called “range of accuracy” and “domain of applicability” of the simulation as a whole, and therefore its constituent parts.

In some cases, like Csma, it can be fairly easy to determine what is not modeled. By reading the model description (csma.h) you can find that there is no collision detection in the CSMA model and decide on how applicable its use will be in your simulation or what caveats you may want to include with your results. In other cases, it can be quite easy to configure behaviors that might not agree with any reality you can go out and buy. It will prove worthwhile to spend some time investigating a few such instances, and how easily you can swerve outside the bounds of reality in your simulations.

As you have seen, ns-3 provides Attributes which a user can easily set to change model behavior. Consider two of the Attributes of the CsmaNetDevice: Mtu and EncapsulationMode. The Mtu attribute indicates the Maximum Transmission Unit to the device. This is the size of the largest Protocol Data Unit (PDU) that the device can send.

The MTU defaults to 1500 bytes in the CsmaNetDevice. This default corresponds to a number found in RFC 894, “A Standard for the Transmission of IP Datagrams over Ethernet Networks.” The number is actually derived from the maximum packet size for 10Base5 (full-spec Ethernet) networks – 1518 bytes. If you subtract the DIX encapsulation overhead for Ethernet packets (18 bytes) you will end up with a maximum possible data size (MTU) of 1500 bytes. One can also find that the MTU for IEEE 802.3 networks is 1492 bytes. This is because LLC/SNAP encapsulation adds an extra eight bytes of overhead to the packet. In both cases, the underlying hardware can only send 1518 bytes, but the data size is different.

In order to set the encapsulation mode, the CsmaNetDevice provides an Attribute called EncapsulationMode which can take on the values Dix or Llc. These correspond to Ethernet and LLC/SNAP framing respectively.

If one leaves the Mtu at 1500 bytes and changes the encapsulation mode to Llc, the result will be a network that encapsulates 1500 byte PDUs with LLC/SNAP framing resulting in packets of 1526 bytes, which would be illegal in many networks, since they can transmit a maximum of 1518 bytes per packet. This would most likely result in a simulation that quite subtly does not reflect the reality you might be expecting.

Just to complicate the picture, there exist jumbo frames (1500 < MTU <= 9000 bytes) and super-jumbo (MTU > 9000 bytes) frames that are not officially sanctioned by IEEE but are available in some high-speed (Gigabit) networks and NICs. One could leave the encapsulation mode set to Dix, and set the Mtu Attribute on a CsmaNetDevice to 64000 bytes – even though an associated CsmaChannel DataRate was set at 10 megabits per second. This would essentially model an Ethernet switch made out of vampire-tapped 1980s-style 10Base5 networks that support super-jumbo datagrams. This is certainly not something that was ever made, nor is likely to ever be made, but it is quite easy for you to configure.

In the previous example, you used the command line to create a simulation that had 100 Csma nodes. You could have just as easily created a simulation with 500 nodes. If you were actually modeling that 10Base5 vampire-tap network, the maximum length of a full-spec Ethernet cable is 500 meters, with a minimum tap spacing of 2.5 meters. That means there could only be 200 taps on a real network. You could have quite easily built an illegal network in that way as well. This may or may not result in a meaningful simulation depending on what you are trying to model.

Similar situations can occur in many places in ns-3 and in any simulator. For example, you may be able to position nodes in such a way that they occupy the same space at the same time, or you may be able to configure amplifiers or noise levels that violate the basic laws of physics.

ns-3 generally favors flexibility, and many models will allow freely setting Attributes without trying to enforce any arbitrary consistency or particular underlying spec.

The thing to take home from this is that ns-3 is going to provide a super-flexible base for you to experiment with. It is up to you to understand what you are asking the system to do and to make sure that the simulations you create have some meaning and some connection with a reality defined by you.

This document was generated on April 21, 2010 using texi2html 1.82.