Table Of Contents
Table Of Contents

Clock Model

Overview

In most communication network simulations, time is simply modeled as a global quantity. All components of the network share the same time throughout the simulation independently of where they are physically located or how they are logically connected to the network.

In contast, in time sensitive networking, the bookkeeping of time is an essential part, which should be explicitly simulated independently of the underlying global time. The reason is that the differences among the local time of the communication network components significantly affects the simulation results.

In such simulations, hardware clocks are simulated on their own, and communication protocols don’t rely on the global value of simulation time, which is in fact unknown in reality, but on the value of their own clocks. With having hardware clocks modeled, it’s also often required to use various time synhronization protocols, because clocks tend to drift over time and communication protocols rely on the precision of the clocks they are using.

In INET, the clock model is a completely optional feature, which has no effect on the simulation performance when disabled. Even if the feature is enabled, the usage of clock modules by communication protocols and applications is still optional, and enabling the feature has negligible performance hit when not in use. For testing that the mere usage of a clock has no effect on the simulation results, INET also includes an ideal clock mechanism.

Clocks

Clocks are implemented as modules, and are used by other modules via direct C++ method calls. Clock modules implement the IClock module interface and the corresponding IClock C++ interface.

The C++ interface provides an API similar to the standard OMNeT++ simulation time based scheduling mechanism, but it relies on the underlying clock implementation for (re)scheduling events according to the clock. These events are transparently scheduled for the client module, and they will be delivered to it when the clock timers expire.

The clock API uses the clock time instead of the simulation time as arguments and return values. The interface contains functions such as getClockTime(), scheduleClockEventAt(), scheduleClockEventAfter(), cancelClockEvent().

INET contains optional clock modules (not used by default) at the network node and the network interface levels. The following clock models are available:

  • IdealClock: clock time is identical to the simulation time.

  • OscillatorBasedClock: clock time is the number of oscillator ticks multiplied by the nominal tick length.

  • SettableClock: a clock which can be set to a different clock time.

Clock Time

In order to avoid confusing the simulation time (which is basically unknown to communication protocols and hardware elements) with the clock time maintained by hardware clocks, INET introduces a new C++ type called the ClockTime.

This type is pretty much the same as the default SimTime, but the two types cannot be implicitly converted into each other. This approach prevents accidentally using clock time where simulation time is needed, and vice versa. Simlarly to how simtime_t is an alias for SimTime, INET also introduces the clocktime_t alias for ClockTime type.

For the explicit conversion between clock time and simulation time, one can use the CLOCKTIME_AS_SIMTIME and the SIMTIME_AS_CLOCKTIME C++ macros. Note that these macros don’t change the numerical value, they simply convert between the C++ types.

When the actual clock time is used by a clock, the value may be rounded according to the clock granularity and rounding mode (e.g. OscillatorBasedClock). For example, when a clock with a us granularity is instructed to wait for 100 ns, while its oscillator is right in the middle of its ticking period, it may actually wait for the next tick to happen to start the timer, and wait another tick to happen to account for the requested wait time interval.

Oscillators

The clock interface is quite general in the sense that it allows many different ways to implement it. Nevertheless, the most common way is to use an oscillator based clock model.

An oscillator efficiently models the periodic generation of ticks that are usually counted by a clock module. The tick period is not necessarily constant, it can change over time. Oscillators implement the IOscillator module interface and the corresponding IOscillator C++ interface.

The following oscillator models are available:

Clock Users

The easiest way to use a clock in applications and communication protocols is to add a clockModule parameter that specifies where the clock module can be found. Then the C++ user module should be simply derived from either ClockUserModuleBase or the parameterizable ClockUserModuleMixin base classes. The clock can be used via the inherited clock related methods or through the methods of the IClock C++ interface on the inherited clock field.

Clock Events

The clock model requires the use of a specific C++ class called ClockEvent to schedule clock timers. It’s also allowed to derive new C++ classes from ClockEvent if necessary. In any case, clock events must be scheduled and canceled via the IClock C++ interface to operate properly.

Controlling Clocks According to a Scenario

In order to support the simulation of specific scenarios, where the clock time or the oscillator drift must be changed according to a predefined script, INET provides clocks and oscillators that implement the interface required by the ScenarioManager module. This allows the user to update the clock and oscillator state from the ScenarioManager XML script and to also mix these operations with many other supported operations.

For example, the SettableClock model supports setting the clock time and also to optionally reset the oscillator at a specific moment of simulation time as follows:

<set-clock at="10 s" module="server.clock" time="1.2 s" reset-oscillator="true"/>

The above example means that the clock time of the server node’s clock will be set to 1.2 seconds when the simulation time reaches 10 seconds, and the clock’s oscillator will restart its duty cycle.

For another example, the ConstantDriftOscillator supports changing the state of the oscillator with the following command:

<set-oscillator at="10 us" module="server.clock.oscillator" drift-rate="42 ppm" tick-offset="1 us"/>

This example simultaneously changes the drift rate and the tick offset of the oscillator in the server node’s clock.

Table Of Contents

Previous
MAC Protocols for Wireless Sensor Networks
Next
The Physical Layer