Escolar Documentos
Profissional Documentos
Cultura Documentos
channel is sensed idle. When it reaches zero, the packet is collisions and causes low overhead. A single channel is
transmitted and the station waits for an acknowledgement adopted, Tx-Rx data rates are assumed to be identical, and
(ACK). If the ACK is not received within a specific period, channel access is based on a Time Division Multiple Access
the station will invoke backoff and retransmission procedure (TDMA) scheme described below. The protocol uses the
using a larger window. An additional RTS/CTS (Request To POLL, NO_DATA, and STATUS control packets, with
Send/Clear To Send) handshake scheme is defined to deal transmission duration tPOLL, tNO_DATA, and tSTATUS, respectively.
with hidden stations. A STATUS packet is marked as ACK or NACK according to
In EDCA, the QoS support is realized with the introduction the specific case. The transmission duration of a DATA
of Access Categories (ACs). The packets are delivered packet is tDATA and the propagation delay is tPROP_DELAY. The
through multiple backoff instances within one station, each possible polling events are:
backoff instance parameterized with AC-specific parameters. - The AP polls an inactive station: The AP sends POLL to
In every station, there are four packet buffers corresponding to the station at time t. The station responds with a NO_DATA
the four ACs. The eight possible User Priorities (UPs) packet, which is received by the AP at
assigned to the generated traffic are mapped to the four ACs. t t POLL t NO_DATA 2t PROP_DELAY . Then, a new poll is
This makes traffic differentiation possible. Each AC within
the stations contends for access independently and starts initiated.
backoff after detecting the channel idle for an AIFS. In order - The AP polls an active station: The AP sends POLL to the
to favor higher priority traffic, higher ACs are assigned lower station at time t. The station replies with a STATUS packet
AIFS values. Furthermore, the higher the AC the smaller its marked as ACK, which carries the destination address and the
contention window. Thus, a high priority packet will probably size of the following DATA packet. Then, the polled station
choose a smaller backoff, increasing its chances to “win” the starts transmitting the DATA packet directly to the
channel contention. destination. Upon reception, the destination broadcasts a
This model provides only minimal QoS. The backoff STATUS packet marked as ACK. Otherwise, if the reception
procedure leads to waste of bandwidth and the hidden stations fails, but the station had successfully identified the source’s
cause collisions despite of the backoff mechanism. The STATUS packet whereby it realized that the following DATA
RTS/CTS handshake limits this problem, however, it increases packet was destined to it, it responds with a STATUS packet
the overhead. Some approaches that enhance EDCA can be marked as NACK. The transmission of a NACK is not wasted
found in [9-11]. An analysis on the performance limits caused time, since either way the stations had to wait for a possible
by EDCA overhead can be found in [12]. EDCA definitely ACK. A new poll is initiated at
enhances DCF with QoS, however, it is shown that it can t t POLL t DATA 2t STATUS 4t PROP_DELAY . We consider
actually serve only limited traffic of low QoS demands. Thus, variable DATA packet size, thus, tDATA is not static.
we propose POAP, which greatly reduces overhead and Specifically, tDATA depends on the size of the currently
optimizes the priority model, providing stricter QoS and transmitted packet.
higher performance. - The communication fails: In case the station does not
When HCCA is also implemented in HCF, the 802.11e receive the POLL packet, the polling fails. The AP has to wait
super-frame is divided into HCCA and EDCA periods. The for the maximum polling cycle before proceeding to a new
stations that want to transmit real-time data with guaranteed poll, since it has to be certain that it will not collide with a
QoS ask the Hybrid Coordinator (HC) for resource possible transmission. When the POLL packet is received by
reservation. The HC schedules transmissions during the the polled station, but then the AP fails to receive any
HCCA periods and grants the stations with channel access feedback, it waits for the maximum polling cycle similarly to
accordingly. HCCA is a complementary protocol in the HCF the previous case. The duration of the maximum polling cycle
scheme. It is not a standalone access method and according to is t POLL t MAX_DATA 2t STATUS 4t PROP_DELAY , where
the 802.11e standard it is only supposed to operate combined
with EDCA. Thus, the direct comparison of HCCA with tMAX_DATA is the duration of the largest allowed DATA packet.
POAP has no meaning and is actually not feasible. POAP and At the end of this cycle, it is certain that the medium is idle.
HCCA are different kinds of protocols. POAP tries to serve When such a communication failure occurs, the AP lowers the
effectively mixed-type traffic, while HCCA cannot even probability to choose this station in the new polling procedure
participate in such a simulation scenario, since it can only assuming a bad link between them. However, it is most likely
handle real-time traffic. For these reasons, HCCA is not taken that the AP will receive some feedback either from the polled
into account when evaluating the POAP protocol. or the destination station.
This scheme provides efficient feedback and low overhead.
The purpose of the control packets is to keep the concerned
III. THE POAP PROTOCOL stations informed of the network status and minimize the idle
intervals. The AP needs to monitor the transmissions so that it
According to POAP, the AP polls the stations to give them
can proceed to the next poll right after the completion of a
permission to transmit. The polling scheme eliminates
communication. Thus, it has to be aware of the actual duration
of the polling cycle. In order to gain this knowledge, the AP buffer selection, the station chooses to send the earliest
just has to detect the NO_DATA packet or the STATUS ACK generated packet in it.
packet, which contains the duration of the following data Before the AP decides which station to poll, it has to be
transmission, or the DATA packet from the polled station or well informed of their buffers' status. Thus, we exploit the
the STATUS ACK-NACK packet from the destination station. ACK and NACK messages, which are already useful.
Actually, when POLL is successfully received, then it is most Specifically, the STATUS packet apart from acknowledging
likely that the AP will obtain the necessary feedback. receptions, it also carries its source's priority score, which is
Moreover, it should be mentioned that despite the fact that an indication of the status of the station's buffered traffic. The
each station is supposed to send a single DATA packet per priority score depends on the priority and the load of each
transmission, it is possible to have multiple successive data buffer. For station j, the priority score is:
PS [ j ] k 0 p[k ] b[k ] . So, the AP examines every
packets destined to the same station with total duration no 3
M 1
for PPR and PB, respectively. Obviously, when in the network PT [ j ] [ j ] l 0
[l ] . The non-normalized final
configuration the purpose is to extendedly favor high priority
probability of polling station j is:
traffic, then WPR is set to a high value compared to WB,
otherwise, if the configuration should be able to efficiently PPOLL [ j ] WPR PP [ j ] WT PT [ j ] , where WT (default
serve highly loaded stations, then the value of WB is raised. 1) is the weight of the contribution of the PT factor.
The default values have resulted from the actual meaning of Obviously, a station that has not been polled for a long time
the parameters and tests which have shown that when the has a high PT value, so its polling probability increases. A
priority weight is three times higher than the buffer load high WT value provides extended fairness among stations,
weight, then the buffer choice probability ensures combination however, this way traffic differentiation fades. If the examined
of efficient traffic differentiation and relatively low packet station j is the AP, then its non-normalized final access
delays for all buffers in most network conditions. We use the probability is multiplied by the factor WAP (default 10), so that
values 6 and 2 rather than 3 and 1, because value 1 is assigned it has clearly higher access chances. Lastly, the AP chooses a
to the weight WT which is introduced later. The non- station to be polled according to each one's normalized polling
normalized choice probability for buffer i is: probability, which is for station j:
P[i] WPR PPR [i] WB PB [i] . As it was mentioned
M 1
PPOLL [ j ] l 0
PPOLL [l ] .
above, when WPR is high compared to WB, it is most probable
that a high priority buffer will be chosen for transmission. On Regarding the overhead caused by POAP, it would be
the other hand, if WB is increased, it is more probable to interesting to have a quantitative comparison with EDCA.
choose a packet from a highly loaded buffer. The normalized Thus, we calculate the duration of a communication without
counting in the data packet transmission time, by considering
3
choice probability equals to: P[i ] k 0
P[k ] . After the the default parameter values for EDCA (AIFS duration, CW
TABLE I. CHARACTERISTICS OF THE TRAFFIC TYPES USED IN THE SIMULATION Voice POAP
SCENARIO
Average Packet Delay (ms) EDCA
Packet
Packet Packet
Inter- On/Off Data
Traffic Data Delay User
Coding arrival Periods Bit
Type Size Bound Priority
Time (sec) Rate
(bytes) (ms) 15
(ms)
~1200
TCP Expo. Always
- 1500 Kbps 60000 1 10
Flow 10 (mean) On
(VBR)
Expo.
32 5
G.726 (mean)
Voice 80 20 Kbps 75 6
(ADPCM) On: 1.5
(CBR)
Off: 1.8
600
Live H.264 Always 0.08
750 10 Kbps 200 5
Video [CIF-20fps] On 0.06 800
(CBR) 0.04 600
0.02 400
Packet Loss Rate 0 200 Voice Throughput (kbps)
duration, RTS size, CTS size, ACK size), assuming that this is
the first transmission attempt and that data rate is 36 Mbps. Fig. 1. Voice: Average packet delay and loss rate versus voice throughput
Concerning POAP, the POLL packet size is 80 bits plus
physical header and the STATUS packet size is 160 bits plus Video
POAP
physical header. In such a case, the time interval while no real Average Packet Delay (ms) EDCA
data is transmitted during an EDCA communication is on the
average equal to 168 μs for the lowest priority traffic and 69
120
μs for the highest priority traffic, while for POAP it is 27.5 μs.
100
Obviously, POAP causes significantly lower overhead than
80
EDCA, even when no collisions and retransmission attempts
are considered for the EDCA scheme. 60
40
20
TCP Total
POAP POAP
Average Packet Delay (ms) EDCA Average Bit Delay (ms) EDCA
8000 0.3
0.25
6000
0.2
4000 0.15
0.1
2000
0.05
Fig. 3. TCP: Average packet delay and loss rate versus TCP throughput Fig. 4. Total Traffic: Average bit delay and loss rate versus total throughput