Читать книгу Wireless Connectivity - Petar Popovski - Страница 25

The system design presented above is not a proper one that can thrive in real life, but rather a sketch of a system that works reasonably well under the assumed collision model. Let us first define Basil's cell to be the area around the base station Basil in which a terminal is in the communication range of Basil. Using our simple collision model, the cell is a circular area.

We can now state that a condition for correct operation of the described TDMA scheme is that any terminal connected to Basil should remain in Basil's cell until Basil decides to terminate the connection. This condition is somewhat strange with respect to the way a link is terminated, since it does not consider the wishes of the terminal. In other words, Basil may decide to terminate the connection to Walt, although Walt may have more data to send in the uplink. However, this is not critical for the overall system operation, as Basil can continue to use the same TDMA structure, substituting Walt with another terminal. What is critical is the case in which Walt has an active link with Basil and Walt walks out of Basil's cell. With the protocol specified above, this leads to a rather fatal system error: the slot allocated to Walt will remain unused forever, as Basil has not terminated the connection and the slot for Walt stays reserved, potentially forever. A practical fix to this situation could be to introduce a certain timeout mechanism: if Basil does not hear from Walt for a certain time and several consecutive time slots allocated for uplink transmission to Walt are silent, then Basil considers Walt to be out of the network and makes Walt's logical channel available to another terminal.

This is still not sufficient to ensure a system design that is robust in practice. Take the following situation: Walt walks temporarily out of reach of Basil but he is back after the timeout has passed. Now Walt does not know that his slot has been allocated to someone else, which may lead to collision in the uplink transmissions made by Walt. The system design can be further patched in different ways in order to deal with this challenge. One solution is that Walt also uses a timeout mechanism, such that if Walt does not send anything to Basil for a time longer than the timeout, then both Basil and Walt claim the link to be non-existent. With this, Walt now knows that he needs to go again through the link establishment procedure.

Alas, this patch is still not sufficient. Recall that the collision model is only a model of reality, but does not fully grasp the practical conditions. One such practical condition is that, even in the absence of collision, the packet is not always received correctly by a receiver that is in the communication range. For example, several consecutive transmissions of Walt may be received incorrectly by Basil due to random noise. In such a situation, Basil starts the timeout for deciding link termination, but Walt does not. This can lead to inconsistent perception of the link between them, since Basil thinks the link is terminated, but Walt thinks the opposite. Yet another patch to the system design can be to use a mechanism based on two-way transmissions between Basil and Walt to check if the link is alive.

We could largely broaden this discussion by spotting other practical deficiencies and finding out suitable patches to the system design. The objective here is not to make a full real-life protocol, but rather illustrate how a simple protocol specification can operate under certain assumptions. However, this protocol needs to be enriched in order to be robust to other practical issues, even for ones that have a very low probability of occurrence.