Because there are usually more than one service connection between a BS and an MSS, we have to consider these connections as a whole. From Fig. 4, we know that if there is more than one connection, the total common free time will decrease, because the total common free time is the common periods of free time among all connections. If we can reduce the number of listening windows in any connection, we can have more common free time among connections, thus have more sleep time to save power. In this thesis, we propose two energy saving schemes to increase the length of common free time and to enhance the energy saving of sleep mode operation. Our schemes were designed for an environment that has both power saving classes of type I and type II connections. In our schemes, we didn’t consider the power saving classes of type III, which is for multicast connections, since we focused on the unicast connections only.
The first proposed scheme is called Longer Common Free Time (LCFT). Because the power saving classes of type II (for UGS, RT-VR) is time-sensitive, we only modify the operation of the power saving classes of type I to have more common free time. For type I connections, the MSS wakes up to listen the traffic indication message at each listening window. The MSS returns to sleep mode again when there is no buffered data in the BS. The
basic idea of our LCFT scheme to removes the listening windows from the power saving classes of type I connections and the traffic indication messages of power saving classes of type I will be handled during the listening windows of power saving classes of type II connections. The reason to do so is because the power saving classes of type I is for connections of BE and NRT-VR, which are time-insensitive.
Fig. 6. Original scheme of power saving classes
14 14
Fig. 7. Proposed method of power saving classes
In Fig. 6, there are three listening windows in the type I connection. The MSS has to wake three times to listen the traffic indication message. If there is only one data coming at time t, in the first and second listening windows, the MSS beeds to wake up and then return to sleep right away. We can have longer common free time if we can keep sleeping at these two listening windows.
In Fig. 7, the proposed LCFT method removes the listening windows from the power saving classes of type 1 connection and the traffic indication message transmitted by the BS will be handled during the listening windows of type II connection. For example, in Fig. 7, the incoming traffic at time t will be handled during the third listening window of type II
connection. The advantage of our LCFT scheme is that for type I connections, the MSS doesn’t need to wake up at all to listen the traffic indication messages from the BS. This method reduces the effect of type I connections on common free time, and the MSS shall have longer free time to enter the sleep mode. On the other hand, the method uses the periodic characteristic of type II connections to read type I traffic indication messages if any. Note that the type II connection wakes up in a fixed period, so the delay of type I connections can be bounded. In summary, this method eliminates of the listening windows of type I connections to save more power while type I connections have bounded delays.
Besides the LCFT, we combine the idea of [10] to enhance the LCFT scheme. We called it enhanced LCFT (E-LCFT). This E-LCFT scheme groups several type II packets in one connection into a single frame for transmission to reduce the number of frames that need to transmit packets from BS to MSS. In this way, the sleep periods of this connection can be increased. As a result, the MSS can have more common free time among different
connections to enter the sleep mode and save more power. The proposed two schemes, LCFT and E-LCFT are compatible with the original IEEE 802.16e standard in terms of no change of MSSs and no change of the communication mechanism between BS and MSS. The only requirement is that the BS needs to be aware of LCFT and E-LCFT in order to set appropriate values of Tmin and Tmax. For the LCFT, assume a type I connection of the MSS sends a request of sleep mode operation to the BS. Then, if the BS permits the request, it will set the
16 16
parameter Tinit to a very large number. For this type I connection, the MSS will not wake up periodically to listen to the traffic indication message. If there are buffered packets for the type I connections in the BS, the BS can transmit the packets to the MSS via the frames of type II connections. If the data size of power saving classes of type I is small enough, we can use the unused frame space of a type II connection to transmit. Otherwise, the BS will transmit the data to the MSS in the next frame and the MSS will stay awake to receive the data. For the E-LCFT, the BS may group several type II packets that are to be sent in separate frames, into a single frame for transmitting later. In this situation, the BS only needs to adjust Tmin to allow the MSS to sleep longer. Again, for E-LCFT to work, the only requirement is the BS needs to be aware of E-LCFT. No other changes are necessary in the MSS or the
communications between BS and MSS.
Our schemes are designed for an environment with both power saving classes of type I and type II connections. If there are many connections of type I and less connections of type II, our schemes may not achieve a good performance. Because if too many buffered data have to be transmitted with type II frames, the average packet delay of type I connections will be extended and the advantage of using unused frame space of type II connections will not work well.