802.11跨层优化

II. INTEGRATED APPLICATION MAC MODELING FOR IEEE
802.11 WLAN A. Application aware MAC layer modeling
As discussed in [2], we model the backoff process by a two dimensional Markov chain denoted as (s(t), b(t)). Here s(t) denotes the backoff stage and b(t) denotes the backoff count value. In order to incorporate the packet arrival from the application (video) layer, we include a new state, referred to as “idle” state. The transition to and from this state is dependent on the arrival rate of the video streaming application. If there is no packet to send or the MAC queue is empty, the system will stay in the idle state. Whenever a packet arrives from the application layer, the 802.11 MAC will follow the backoff procedure as described in [1]. Once the MAC has transmitted the current frame successfully from its queue (i.e. received ACK frame from the receiver), it will reset its retransmission count and process the next frame in the queue in case this is available. After transmitting a frame successfully, the MAC enters the post backoff stage wherein it chooses a random value from the first backoff stage. This ensures that between two packet transmissions, there is a backoff stage. The backoff stage was already considered in several papers, including [3][4][7]. When the backoff counter is equal to 0, the MAC sends the packet to the channel. In this case, when collisions occur, the exponential backoff algorithm will take place and the contention window will be doubled [1]. The Markov model described in [2][3] is able to handle the backoff procedure, however, it assumes that the retransmission limit at the MAC is infinite.
Nevertheless, for real-time video transmission applications operating under stringent delay constraints ranging between 200ms up to several seconds, the retransmission limit needs to be optimally adjusted. Our previous adaptation of the MAC retransmission limits considers only a simplistic MAC model, which differs considerably from the DCF mechanism employed in 802.11 WLAN MAC cards. Hence, our algorithm needs to be modified to consider the backoff stage, which has significant impact on the delay incurred at the application layer.
Our proposed application-aware MAC model is depicted in Figure 1. From the figure, the one step transitions are given by the following equations:
P{idle|idle}=π0 (1)
P{0,0|idle}=1 π0 (2)
P{i,k|i,k+1}=1 q,0≤i≤m,0≤k≤Wi-2
(3)
P{i,k|i,k}=q,0≤i≤m,0<k≤Wi-1 (4)
qcπ1
i=, 0≤k≤Wi-1 , 0 W1
(5) P{i+1,k|i,0}=
qc,1 ≤i≤m, 0≤k≤Wi-1 W1
i=,0≤k≤Wi-1 (1 qc)π1, 0
P{0,k|i,0}= 1-qc,0<i<m, 0≤k≤Wi-1 (6)
i=m,≤k≤Wi-1 1, 0
P{idle|0,0)=1 π1 (7)
In the above equations, π0 is the probability that the MAC queue is empty. π1 is the probability that the system is busy (i.e. has packets to send), but has not yet performed the post backoff process. q is the probability that the given slot is active (busy). qc is the collision probability given that the slot is active and m is the backoff stage. The window sizes at the backoff stage are represented by W0,..Wm.
Equation (1) describes the case when there is no packet in the queue. In this case, the MAC stays in the same idle state with probability π0. Equation (2) represents the case when the
MAC has a packet to send, but it will not go into the backoff procedure because the post backoff was already performed and the medium was idle for longer than Distributed Inter-Frame Space (DIFS) time. Equations (3) and (4) represent the backoff counter decrementing process. If the sensed slot is idle, the MAC will decrement its backoff value by one. If the medium is busy, the MAC will freeze its backoff value. Collisions occur when two WSTAs (wireless stations) try to send out packets in the same slot. This is represented by Equation (5). Whenever a collision occurs, the MAC doubles the size of the backoff
window and chooses a new backoff counter uniformly in the
Figure 1: Markov Chain Model for IEEE 802.11 Backoff Process that
range of 0 to the backoff window minus one, i.e. [0,Wi 1]. incorporates the Application Layer Arrival.
Note that the first part of Equation (5) has an extra term π1,
corresponding to the case that the WSTA’s queue is non empty
application layer to achieve the best video ...Zhang, “An integrated scheme for automatic video...Cross-layer Optimized Wireless Multimedia ...
cross-layer design are demonstrated for video ... or the application layer can adapt its ...to reach optimized usage of the wireless channel....
Integrated MAC and Routing Protocol for Wireless ...solutions optimized for specic application classes. ... cross-layer integration, MAC, power-saving mode...
Proxy-driven ratedistortion optimized video ... MAC layer FEC/ARQ, application layer FEC and ...Following the discussion of channel modeling in ...
Cross-layer optimization for OFDM wireless ...provisioning when application performance is the key...when they are optimized for the sum of ...
cross-layer optimized system to maximize the ...layer are integrated for wireless ad hoc networks... video encoder resides in the application layer. ...
A cross-layer framework for wireless LAN QoS ...They can be assigned directly by the application ...from the performance modeling to a compact policy...
wireless home networks as an application example. ... utilizing the rate-distortion optimized framework ...The bene?t of cross-layer video rate adaptation...
Video Transmission over Wireless LANs Pierre Ferr?...and can also be integrated onto handheld devices....Layer Application Transport Network Link/MAC Packet...
over Wireless Networks with Heterogeneous L_专业...optimized rate allocation among multiple video ...MAC layer and the applicationlayer video rate ...
热门文档