**3. Simulation results**

This section evaluated the power consumption of an MSS in terms of the number of listen time slots and status transitions. The QoS requirements of *A*, *B*, *C*, and *D* are listed in Table 2. Both connection types *A* and *B* are VoIP connections. Both connection types *C* and *D* are video streaming connections. The first four connection types on the top half of the list are real-time connections that do not consider the tolerated jitter, and the last four connection types are the same as the first four connection types, but with constrained tolerated jitter. The total energy of an MSS is 1,000,000 units. We compare our proposed SSS algorithm with the Naïve approach without optimizations and the AS approach (Tsao & Chen, 2008). The Naïve approach implies that each connection associates with its preferred type of powersaving class and parameters, and minimizs that packet delay and power consumption for that single connection.

Fig. 8 shows the operation time and energy usage of an MSS by applying three different scheduling schemes in the different connection types with a varied number of connections without the jitter constraints. In the Naïve approach, the energy usage increases faster than the other two approaches. Because the Naïve approach does not consider the optimization of packet scheduling, it results in the enormous energy consumption in status transitions. The energy usage in the AS approach performs the same as our SSS approach when there is

A QoS Guaranteed Energy-Efficient Scheduling for IEEE 802.16e 43

(b)

(c)

only one connection in an MSS. This is because the two approaches maximize the delay of packets scheduling and schedule the packets into minimal listen periods. However, since we consider status transitions in scheduling the packets, our SSS approach chooses successive frames in scheduling the packets and reducing the number of status transitions. When more connections take into account the scheduling, our SSS approach reduces the number of status transitions by successive scheduling. In other words, while successive time slots are scheduled with packets, they do not place the status transitions in the time slots. Thus, our SSS algorithm saves energy and prolongs the operation time in an MSS.


Table 2. QoS parameters of four real-time connections.

(a)

42 Mobile Networks

only one connection in an MSS. This is because the two approaches maximize the delay of packets scheduling and schedule the packets into minimal listen periods. However, since we consider status transitions in scheduling the packets, our SSS approach chooses successive frames in scheduling the packets and reducing the number of status transitions. When more connections take into account the scheduling, our SSS approach reduces the number of status transitions by successive scheduling. In other words, while successive time slots are scheduled with packets, they do not place the status transitions in the time slots. Thus, our

> Interval of packets arrival (ms)

> > (a)

*A* UGS 32 10 50 ∞ *B* UGS 128 10 50 ∞ *C* RT-VR 512 30 100 ∞ *D* RT-VR 1024 30 100 ∞ *A'* UGS 32 10 50 10 *B'* UGS 128 10 50 10 *C'* RT-VR 512 30 100 20 *D'* RT-VR 1024 30 100 20

Delay constraint (ms)

Tolerated jitter (ms)

SSS algorithm saves energy and prolongs the operation time in an MSS.

Packets size (bytes)

Table 2. QoS parameters of four real-time connections.

Service type of QoS

(b)

(c)

A QoS Guaranteed Energy-Efficient Scheduling for IEEE 802.16e 45

(a)

(b)

Fig. 8. The operation time and energy usage of an MSS for three schemes with four connection types with a varied number of connections: (a) connection type *A*, (b) connection type *A*+*B*, (c) connection type *A*+*B*+*C*, and (d) connection type *A*+*B*+*C*+*D*.

Fig. 9 shows the average energy efficiency of an MSS by applying three different scheduling schemes for different connection types with a varied number of connections without the jitter constraints. We defined *Etrans* as the energy usage for the packet transmission of an MSS during a time period *T*; *Etotal* represents the total energy usage in an MSS during *T*. The average energy efficiency (*AEE*) for an MSS during *T* can be represented as follows:

$$\text{AEE} = \text{E}\_{\text{trans}} \int \text{E}\_{\text{total}} \tag{2}$$

In the Naïve approach, the average energy efficiency is lower than the other two approaches. This is because the Naïve approach processes packets immediately when they arrive, so number of status transitions increase enormously. The energy for status transitions reduce the energy usage for packet transmission from the total energy usage in an MSS. In our SSS algorithm, the average energy efficiency performed the same as the AS approach, where there is only one connection in an MSS. The reason for this is the same as the previous simulation matrix. When there is only one connection in an MSS, the two approaches maximize the delay in packet scheduling and schedules the packets into their minimal listen periods without violating the delay constraints. Thus, the number of status transitions is the same. However, the average energy efficiency in our SSS approach grows up when the number of connections increases. This is because the packets are scheduled more successively when the packets are small, and the number of connections grows large under our proposed algorithm. Fig. 9(c) and (d) reveal that, when the transmission loading encounters a bottleneck, the average energy efficiency stops increasing.

44 Mobile Networks

(d)

connection types with a varied number of connections: (a) connection type *A*, (b) connection

Fig. 9 shows the average energy efficiency of an MSS by applying three different scheduling schemes for different connection types with a varied number of connections without the jitter constraints. We defined *Etrans* as the energy usage for the packet transmission of an MSS during a time period *T*; *Etotal* represents the total energy usage in an MSS during *T*. The

In the Naïve approach, the average energy efficiency is lower than the other two approaches. This is because the Naïve approach processes packets immediately when they arrive, so number of status transitions increase enormously. The energy for status transitions reduce the energy usage for packet transmission from the total energy usage in an MSS. In our SSS algorithm, the average energy efficiency performed the same as the AS approach, where there is only one connection in an MSS. The reason for this is the same as the previous simulation matrix. When there is only one connection in an MSS, the two approaches maximize the delay in packet scheduling and schedules the packets into their minimal listen periods without violating the delay constraints. Thus, the number of status transitions is the same. However, the average energy efficiency in our SSS approach grows up when the number of connections increases. This is because the packets are scheduled more successively when the packets are small, and the number of connections grows large under our proposed algorithm. Fig. 9(c) and (d) reveal that, when the transmission loading

AEE= Etrans / Etotal (2)

Fig. 8. The operation time and energy usage of an MSS for three schemes with four

average energy efficiency (*AEE*) for an MSS during *T* can be represented as follows:

type *A*+*B*, (c) connection type *A*+*B*+*C*, and (d) connection type *A*+*B*+*C*+*D*.

encounters a bottleneck, the average energy efficiency stops increasing.

A QoS Guaranteed Energy-Efficient Scheduling for IEEE 802.16e 47

Fig. 10 shows the operation time and energy usage of an MSS by applying three different scheduling schemes for different connection types with a varied numbers of connections with the jitter constraints. The energy usage of different three approaches is higher than the one that does not consider the jitter constraints. The reason for this is that the process is limited

(a)

Fig. 9. The average energy efficiency of an MSS with three schemes and four connection types with a varied number of connections: (a) connection type *A*, (b) connection type *A*+*B*, (c) connection type *A*+*B*+*C*, and (d) connection type *A*+*B*+*C*+*D*.

46 Mobile Networks

(d) Fig. 9. The average energy efficiency of an MSS with three schemes and four connection types with a varied number of connections: (a) connection type *A*, (b) connection type *A*+*B*,

(c) connection type *A*+*B*+*C*, and (d) connection type *A*+*B*+*C*+*D*.

(c)

Fig. 10 shows the operation time and energy usage of an MSS by applying three different scheduling schemes for different connection types with a varied numbers of connections with the jitter constraints. The energy usage of different three approaches is higher than the one that does not consider the jitter constraints. The reason for this is that the process is limited

$$\text{(a)}$$

A QoS Guaranteed Energy-Efficient Scheduling for IEEE 802.16e 49

by the jitter constraints, and the limited scheduling increases the number of status transitions. In our SSS approach, the energy usage is lower than the other two approaches under the same connection types. That is because the more connections gain the more chances to be scheduled

Fig. 11 shows the amount of packet loss of an MSS which applies two different scheduling schemes for different connection types with a varied number of connections with the jitter

(a)

(b)

successively, so the energy consumption of status transitions is reduced.

(c)

Fig. 10. The operation time of an MSS with three schemes and four connection types with a varied number of connections with jitter constraints: (a) connection type *A*', (b) connection type *A*'+*B*', (c) connection type *A*'+*B*'+*C*', (d) connection type *A*'+*B*'+*C*'+*D*'.

48 Mobile Networks

(c)

(d) Fig. 10. The operation time of an MSS with three schemes and four connection types with a varied number of connections with jitter constraints: (a) connection type *A*', (b) connection

type *A*'+*B*', (c) connection type *A*'+*B*'+*C*', (d) connection type *A*'+*B*'+*C*'+*D*'.

by the jitter constraints, and the limited scheduling increases the number of status transitions. In our SSS approach, the energy usage is lower than the other two approaches under the same connection types. That is because the more connections gain the more chances to be scheduled successively, so the energy consumption of status transitions is reduced.

Fig. 11 shows the amount of packet loss of an MSS which applies two different scheduling schemes for different connection types with a varied number of connections with the jitter

$$\bf{(a)}$$

A QoS Guaranteed Energy-Efficient Scheduling for IEEE 802.16e 51

that considers jitter constraints. The scheduler chooses the proper time slots to schedule the

Fig. 12 shows the average energy efficiency of an MSS by applying two different scheduling schemes for different connection types with a varied number of connections with jitter

(a)

(b)

packets in order so as not violate the jitter constraints between each packet.

Fig. 11. The amount of packet loss of an MSS with two schemes and four connection types with a varied number of connections: (a) with 1 connection, (b) with 8 connections, (c) with 16 connections, and (d) with 32 connections.

constraints. We only compare the SSS and the AS approaches, which delay the packets, when processing the scheduling. The amount of packet loss is increased when the packet load is raised. In our SSS approach, the number of packet loss is minimized by the algorithm 50 Mobile Networks

(c)

(d) Fig. 11. The amount of packet loss of an MSS with two schemes and four connection types with a varied number of connections: (a) with 1 connection, (b) with 8 connections, (c) with

constraints. We only compare the SSS and the AS approaches, which delay the packets, when processing the scheduling. The amount of packet loss is increased when the packet load is raised. In our SSS approach, the number of packet loss is minimized by the algorithm

16 connections, and (d) with 32 connections.

that considers jitter constraints. The scheduler chooses the proper time slots to schedule the packets in order so as not violate the jitter constraints between each packet.

Fig. 12 shows the average energy efficiency of an MSS by applying two different scheduling schemes for different connection types with a varied number of connections with jitter

A QoS Guaranteed Energy-Efficient Scheduling for IEEE 802.16e 53

constraints. In this simulation, we only compare the SSS and AS approaches, which delay the packets when processing the scheduling. In our SSS algorithm, the average energy efficient is better than the performance of the AS approach. Due to QoS constraints, the available time slots for scheduling was limited by the delay and jitter constraints. Aside from the energy usage of status transitions, the packets will not be delivered if the scheduling violates the delay and jitter constraints. Meanwhile, the AS does not take the jitter constraints into account when they scheduling the packets. Thus, our SSS approach transmits more packets than the AS, and the average energy efficient in our SSS approach is

An energy-efficient scheduling scheme to improve the energy efficiency and guarantee Quality of Service in IEEE 802.16e was proposed. The previous literature only considers the delay constraint of QoS requirement in one MSS. We first consider both the jitter and delay constraints of QoS requirement to schedulethe real-time connections in one MSS. Our proposed algorithmis to schedule the packet transmission in successively fashion with the minimal interval of listen periods and maximal interval of sleep periods without violating the QoS of all connections in an MSS. Additionally, the successive scheduling of time slots would reduce the number of status transitions between the sleep periods and listen periods. The proposed approach can be adapted to the power-saving class of type III where the length of sleep and listen periods arevariable. Simulation results show that, incomparison with the AS and Naïve schemes, the proposed SSS scheduling algorithm can result in a

Andrews, M.; Qian, L. & Stolyar, A. (2005). Optimal Utility Based Multi-user Throughput

Fang, G.; Dutkiewicz, E.; Sun, Y.; Zhou, J.; Shi, J. & Li, Z. (2006). Improving Mobile Station

Huang, S.-C.; Jan, R.-H. & Chen, C. (2007). Energy Efficient Scheduling with QoS Guarantee

Jang, J.; Han, K. & Choi, S. (2006). Adaptive Power Saving Strategies of IEEE 802.16e Mobile

Allocation Subject to Throughput Constraints, *IEEE INFOCOM*, pp. 2415- 2424,

Energy Efficiency in IEEE 802.16e WMAN by Burst Scheduling, *IEEE International Conference on Global Telecommunications Conference (GLOBECOM'06)*, pp. 1-5, 2006. Han, K. & Choi, S. (2006). Performance Analysis of Sleep Mode Operation in IEEE 802.16e

Mobile Broadband Wireless Access Systems, *IEEE 63rd Vehicular Technology* 

for IEEE 802.16e Broadband Wireless Access Networks, *International Conference on Wireless Communications and Mobile Computing (IWCMC'07)*, pp. 547-552, 2007. IEEE Std 802.16-2004 (2004). IEEE Standard for Local and Metropolitan Area Networks. Part 16: Air Interface for Fixed Broadband Wireless Access Systems, 2004. IEEE Std 802.16e-2005 (2005). IEEE Standard for Local and Metropolitan Area Networks.

Part 16: Air Interface for Fixed and Mobile Broadband Wireless Access Systems,

Broadband Wireless Access, *Asia-Pacific Conference on Communications (APCC'06)*,

significant overall energy saving and can guarantee the delay and jitter QoS.

*Conference (VTC 2006-Spring),* pp. 1141-1145, 2006.

better than the AS.

**4. Conclusion** 

**5. References** 

2005.

2005.

pp. 1-5, 2006.

Fig. 12. The average energy efficiency of an MSS with two schemes and four connection types with a varied number of connections under jitter constraints: (a) connection type *A*', (b) connection type *A*'+*B*', (c) connection type *A*'+*B*'+*C*', and (d) connection type *A*'+*B*'+*C*'+*D*'.

constraints. In this simulation, we only compare the SSS and AS approaches, which delay the packets when processing the scheduling. In our SSS algorithm, the average energy efficient is better than the performance of the AS approach. Due to QoS constraints, the available time slots for scheduling was limited by the delay and jitter constraints. Aside from the energy usage of status transitions, the packets will not be delivered if the scheduling violates the delay and jitter constraints. Meanwhile, the AS does not take the jitter constraints into account when they scheduling the packets. Thus, our SSS approach transmits more packets than the AS, and the average energy efficient in our SSS approach is better than the AS.
