Most studies in heterogeneous network (IEEE 802.11 and IEEE 802.16e) are related with the handoff between IEEE 802.11 and IEEE 802.16e, like [7] and [8].
Nobody study about the coexisting of IEEE 802.11 traffic and IEEE 802.16e traffic in heterogeneous network. In this chapter, we focus on other work about the power management in IEEE 802.11 and IEEE 802.16e.
2.1 Work in IEEE 802.11
For IEEE 802.11, a simple method to conserve more power is having a long listen interval. In this way, mobile station will stay more time in power save mode.
The drawback is that when traffic load is heavy buffered packets in AP will exceed the queue size before mobile station wakeup. To solve this problem, many studies are proposed. In [9] and [10], they propose listen interval adaptation mechanisms in which the mobile station dynamically adapts the duration of listen interval according to the traffic situation. Another problem of power save mode in IEEE 802.11 will appear when the numbers of mobile stations in power save mode is more. Some mobile stations might have no chance to get the buffered packets in AP causing by too many mobile stations wakeup at the same beacon interval. In [11], the authors propose a method to arrange the wakeup schedule for sleeping mobile stations such that the number of wakeup mobile stations in each beacon interval is balanced.
2.2 Work in IEEE 802.16e
For power saving mechanism in IEEE 802.16e, [12] [13] [14] [15] analyze the binary-increasing sleep window of power saving classes of type I by proposing a
mathematical model. In [16], the author also proposes mathematical models to calculate the power efficiency and packet access delay. Unlike [12] [13] [14] [15]
which only considering power saving classes of type I, [16] propose two models for power saving classes of type I and type II. The result in [16] shows power saving classes of type I can get better performance in power saving, but have worse packet delay causing by binary-increasing sleep window.
In [17], the authors propose two scheduling algorithms for sleep mode operations in power saving classes of type II. The first algorithm is periodic on-off scheme for connection which distributes small packets to all OFDM frames group these small packets together without violating the QoS requirement. Figure 2-1 is an example of periodic on-off scheme.
Grouping
Listen Sleep Listen
OFDM frame UL/DL packets Original
Listen Sleep
Figure 2-1 Periodic on-off scheme
The second algorithm is aperiodic on-off scheme. It merges the transmissions of different connections into fewer frames without violating the QoS requirements. It delays the transmission of connections and seeks if having transmission of other connection within the delay constrain. Figure 2-2 is an example of aperiodic on-off scheme. From (b), C2 will delay its transmissions and it finds C1 has transmissions within its delay constrain. From the view of mobile station, mobile station has fewer transmissions in (b). The reason is some transmissions of C2 transmit with C1.
C1 C2
Time C1
C2 MS
MS (a)
(b)
Transmission of C 1 Transmission of C 2 Transmission of MS
Figure 2-2 Aperiodic on-off scheme. (a) Original, (b) aperiodic on-off scheme.
Even if these two algorithms can reduce the times of transmission, there still are some drawbacks. First, they only consider the periodic fixed size traffic which be know before and ignore the variable size traffic like rtPS. The information of queue size of rtPS connection is obtained by polling or piggyback and is know in the run-time. Second, they don’t consider the QoS requirements of power saving classes of type I. Third, the aperiodic on-off scheme is not optimal. In aperiodic on-off scheme, merging of transmission is done per connections. In order to solve the problems, the author of [18] proposes a new transmission merging mechanism.
In [18], there are three parts of this transmission merging mechanism. The first part handles periodic fixed size traffics, like UGS. It will calculate the next merging candidate set. Every frame within the set is optimal solution. The second part handles the variable size traffics, like rtPS, nrtPS and BE. It first calculates the size of next listening window according to the queue size. Then it calculates the maximum size of next sleep window according to the QoS requirement constrain and next listening
window. Finally, from the next merging candidate set and maximum size of sleep window, the third part determines the actual listening window and sleep window.
C1 C2
Step 1 Step 2 MS
Transmission of C 1 Transmission of C 2 Transmission of MS
(a)
(b)
MS C1 C2
Figure 2-3 Merging methods in [17] and [18]. (a) [17] (b) [18]
The merging methods in [17] and [18] are shown in figure 2-3. Figure 2-3 is the worst case of aperiodic on-off scheme. From this figure, we can see the drawback of aperiodic on-off scheme. In aperiodic on-off scheme, it will choose the connection with small delay constrain first. In figure 2-3, the delay constrains of C1 and C2 are 1 and 2. First step, the aperiodic on-off scheme will choose C1 and can’t find any other transmissions within its delay constrain. In this situation, the aperiodic on-off scheme will delay the transmissions of C1 to reach maximum delay shown in figure 2-3 (a).
Second step, it will choose C2 and can’t find any other transmissions within its delay constrain again. Then it does the same behavior like C1. Finally, we can see there is no transmission been merged. Unlike aperiodic on-off scheme in [17], the
transmission merging mechanism in [18] will take all connections into consideration.
From figure 2-3 (b), we can find that the mobile station only needs five transmissions.
More detail about this mechanism will be described in next chapter.