- •Introduction
- •Increasing Demand for Wireless QoS
- •Technical Approach
- •Outline
- •The Indoor Radio Channel
- •Time Variations of Channel Characteristics
- •Orthogonal Frequency Division Multiplexing
- •The 5 GHz Band
- •Interference Calculation
- •Error Probability Analysis
- •Results and Discussion
- •IEEE 802.11
- •IEEE 802.11 Reference Model
- •IEEE 802.11 Architecture and Services
- •Architecture
- •Services
- •802.11a Frame Format
- •Medium Access Control
- •Distributed Coordination Function
- •Collision Avoidance
- •Post-Backoff
- •Recovery Procedure and Retransmissions
- •Fragmentation
- •Hidden Stations and RTS/CTS
- •Synchronization and Beacons
- •Point Coordination Function
- •Contention Free Period and Superframes
- •QoS Support with PCF
- •The 802.11 Standards
- •IEEE 802.11
- •IEEE 802.11a
- •IEEE 802.11b
- •IEEE 802.11c
- •IEEE 802.11d
- •IEEE 802.11e
- •IEEE 802.11f
- •IEEE 802.11g
- •IEEE 802.11h
- •IEEE 802.11i
- •Overview and Introduction
- •Naming Conventions
- •Enhancements of the Legacy 802.11 MAC Protocol
- •Transmission Opportunity
- •Beacon Protection
- •Direct Link
- •Fragmentation
- •Traffic Differentiation, Access Categories, and Priorities
- •EDCF Parameter Sets per AC
- •Minimum Contention Window as Parameter per Access Category
- •Maximum TXOP Duration as Parameter per Access Category
- •Collisions of Frames
- •Other EDCF Parameters per AC that are not Part of 802.11e
- •Retry Counters as Parameter per Access Category
- •Persistence Factor as Parameter per Access Category
- •Traffic Streams
- •Default EDCF Parameter Set per Draft 4.0, Table 20.1
- •Hybrid Coordination Function, Controlled Channel Access
- •Controlled Access Period
- •Improved Efficiency
- •Throughput Improvement: Contention Free Bursts
- •Throughput Improvement: Block Acknowledgement
- •Delay Improvement: Controlled Contention
- •Maximum Achievable Throughput
- •System Saturation Throughput
- •Modifications of Bianchi’s Legacy 802.11 Model
- •Throughput Evaluation for Different EDCF Parameter Sets
- •Lower Priority AC Saturation Throughput
- •Higher Priority AC Saturation Throughput
- •Share of Capacity per Access Category
- •Calculation of Access Priorities from the EDCF Parameters
- •Markov Chain Analysis
- •The Priority Vector
- •Results and Discussion
- •QoS Support with EDCF Contending with Legacy DCF
- •1 EDCF Backoff Entity Against 1 DCF Station
- •Discussion
- •Summary
- •1 EDCF Backoff Entity Against 8 DCF Stations
- •Discussion
- •Summary
- •8 EDCF Backoff Entities Against 8 DCF Stations
- •Discussion
- •Summary
- •Contention Free Bursts
- •Contention Free Bursts and Link Adaptation
- •Simulation Scenario: two Overlapping QBSSs
- •Throughput Results with CFBs
- •Throughput Results with Static PHY mode 1
- •Delay Results with CFBs
- •Conclusion
- •Radio Resource Capture
- •Radio Resource Capture by Hidden Stations
- •Solution
- •Mutual Synchronization across QBSSs and Slotting
- •Evaluation
- •Simulation Results and Discussion
- •Conclusion
- •Prioritized Channel Access in Coexistence Scenarios
- •Saturation Throughput in Coexistence Scenarios
- •MSDU Delivery Delay in Coexistence Scenarios
- •Scenario
- •Simulation Results and Discussion
- •Conclusions about the HCF Controlled Channel Access
- •Summary and Conclusion
- •ETSI BRAN HiperLAN/2
- •Reference Model (Service Model)
- •System Architecture
- •Medium Access Control
- •Interworking Control of ETSI BRAN HiperLAN/2 and IEEE 802.11
- •CCHC Medium Access Control
- •CCHC Scenario
- •CCHC and Legacy 802.11
- •CCHC Working Principle
- •CCHC Frame Structure
- •Requirements for QoS Support
- •Coexistence Control of ETSI BRAN HiperLAN/2 and IEEE 802.11
- •Conventional Solutions to Support Coexistence of WLANs
- •Coexistence as a Game Problem
- •The Game Model
- •Overview
- •The Single Stage Game (SSG) Competition Model
- •The Superframe as SSG
- •Action, Action Space A, Requirements vs. Demands
- •Abstract Representation of QoS
- •Utility
- •Preference and Behavior
- •Payoff, Response and Equilibrium
- •The Multi Stage Game (MSG) Competition Model
- •Estimating the Demands of the Opponent Player
- •Description of the Estimation Method
- •Evaluation
- •Application and Improvements
- •Concluding Remark
- •The Superframe as Single Stage Game
- •The Markov Chain P
- •Illustration and Transition Probabilities
- •Definition of Corresponding States and Transitions
- •Solution of P
- •Collisions of Resource Allocation Attempts
- •Transition Probabilities Expressed with the QoS Demands
- •Average State Durations Expressed with the QoS Demands
- •Result
- •Evaluation
- •Conclusion
- •Definition and Objective of the Nash Equilibrium
- •Bargaining Domain
- •Core Behaviors
- •Available Behaviors
- •Strategies in MSGs
- •Payoff Calculation in the MSGs, Discounting and Patience
- •Static Strategies
- •Definition of Static Resource Allocation Strategies
- •Experimental Results
- •Scenario
- •Discussion
- •Persistent Behavior
- •Rational Behavior
- •Cooperative Behavior
- •Conclusion
- •Dynamic Strategies
- •Cooperation and Punishment
- •Condition for Cooperation
- •Experimental Results
- •Conclusion
- •Conclusions
- •Problem and Selected Method
- •Summary of Results
- •Contributions of this Thesis
- •Further Development and Motivation
- •IEEE 802.11a/e Simulation Tool “WARP2”
- •Model of Offered Traffic and Requirements
- •Table of Symbols
- •List of Figures
- •List of Tables
- •Abbreviations
- •Bibliography
5.1 HCF Contention-based Channel Access |
89 |
In the following, the resulting throughput for backoff entities of the AC with higher priority (EDCF backoff entities) and with legacy priority (legacy DCF stations8) are discussed for different scenarios. It is shown that EDCF backoff entities do not achieve the desired priority over the legacy DCF stations. Two measures to support a better priority over legacy stations are therefore discussed in this context: the PF and the use of EIFS instead of DIFS. Legacy stations can be forced to use EIFS instead of DIFS by using Frame Check Sequences (FCSs) that are different from the DCF (Hiertz, 2002). This is one of the concepts used in the context of interworking of different wireless LANs and discussed in detail in Section 6.2.1.2.
Three different scenarios are examined in the following sections. In Section 5.1.4.1, scenarios with one EDCF backoff entity and one legacy DCF station operating in parallel are discussed. In Section 5.1.4.2, the more problematic case when one EDCF backoff entity operates in parallel to multiple legacy DCF stations is discussed. Finally, in Section 5.1.4.3, scenarios with multiple EDCF backoff entities and multiple DCF stations operating in parallel are discussed.
In all simulated scenarios, all stations can detect each other. Hence, there is no hidden station. All frame bodies are 512 byte long, neither RTS/CTS nor fragmentation is used. Inter-arrival times are negative-exponentially distributed. The 16QAM1/2 PHY mode is used; the radio channel is error free.
5.1.4.11 EDCF Backoff Entity Against 1 DCF Station
Figure 5.20 illustrates the scenario and Figure 5.21 shows the resulting throughput per AC vs. the offered traffic per backoff entity (a-c), and the distribution of backoff delay in saturation (d).
|
using high |
using DIFS, or |
priority with |
||
EDCF |
different |
EIFS instead |
PF |
of DIFS |
|
backoff |
DCF |
|
entity |
|
|
|
|
legacy |
|
|
station |
receiving station
Figure 5.20: Scenario. All stations detect each other. If the two stations transmit at the same time, a collision occurs.
8For legacy 802.11, “DCF station” and “DCF backoff entity” can be used as synonym for each other, because there is one backoff entity per station in legacy 802.11.
90 |
5. Evaluation of IEEE 802.11e with the IEEE 802.11a Physical Layer |
5.1.4.1.1Discussion
Figure 5.21(a) shows the results for the standard configuration, where the EDCF backoff entity operates with the higher priority EDCF parameters defined in Table 5.2, but with the persistence factor set as defined in 802.11e, i.e., PF=2.
(AIFSN=2, CWmin=7, CWmax=1023, PF=2, RetryCounter=7). Note that PF=2 is the only value used in 802.11e, according to draft 4.0 (IEEE 802.11 WG, 2002c). Shown are simulation results and results obtained with the analytical model that is described in the previous section. It can be seen that the EDCF backoff entity achieves a higher throughput than the legacy DCF station, because of the smaller size of the initial contention window, i.e., CWmin.
It is interesting to investigate additional concepts that are not 802.11e conformant, to increase the relative priority of EDCF over legacy DCF. In the following, the influence of the PF and an increase of the interframe space from DIFS to EIFS are evaluated.
Figure 5.21(b) shows the results for scenarios where the EDCF backoff entity operates with the high priority EDCF parameters, now including the PF. The PF is now 1.5 instead of 2. With only two contending backoff entities, a smaller PF is not helpful, as the number of collisions is relatively small. Thus, the results in Figure 5.21(b) do not significantly diverge from the results in Figure 5.21(a). With a small number of collisions, the influence of the PF on the achievable throughput is negligible. Figure 5.21(c) shows the results for scenarios where the legacy DCF station is forced to operate with EIFS instead of DIFS all the time. Now, the priority of the EDCF backoff entity is clearly visible, thanks to the increased interframe space used by the legacy DCF station.
The analytical results in Figure 5.21(a-c) deviate from the simulation results because of the used assumption that the access probability per slot is geometrically distributed. This is not the case with one backoff entity per AC. With one backoff entity per AC, the access probability per slot is uniformly distributed. However, the analytical results show at least the same characteristics of the saturation throughput per AC relative to each other. The analytical model described in the previous section gives the saturation throughput per AC in a shared scenario, which is here the throughput per backoff entity when the offered traffic is high (overload scenario). Instead of illustrating the results as one single point in the figure, they are indicated as maximum achievable throughput when the offered traffic is increased (indicated as line).
Figure 5.21(d) illustrates the Complementary Cumulative Distribution Functions
(CCDFs) of the backoff delay for all three scenarios. The EDCF backoff entity
5.1 HCF Contention-based Channel Access |
91 |
[Mbit/s] |
16 |
PF=2, AIFS=DIFS. |
|
|
DCF sim. |
|
|||
|
14 |
1 EDCF against 1 DCF station. |
|
EDCF sim. |
|
||||
|
|
EDCF analyt. |
|||||||
|
|
|
|
|
|
|
|||
AC |
12 |
|
|
|
|
|
DCF analyt. |
|
|
|
|
|
|
|
|
|
|||
|
|
|
|
|
|
|
|
||
per |
10 |
|
|
|
|
|
|
|
|
8 |
|
|
|
|
|
|
|
|
|
thrp |
|
|
|
|
|
|
|
|
|
6 |
|
|
|
|
|
|
|
|
|
cummulative |
|
|
|
|
|
|
|
|
|
4 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
|
|
|
|
|
|
|
|
0 |
|
|
|
|
|
|
|
|
|
0 |
2 |
4 |
6 |
8 |
10 |
12 |
14 |
16 |
|
|
|
Offered traffic per station [Mbit/s] |
|
|
||||
|
|
|
(a) standard configuration |
|
|
||||
[Mbit/s] |
16 |
|
|
|
|
|
|
|
|
14 |
PF=2, AIFS=DIFS, DCF uses EIFS. |
|
|
|
|||||
|
1 EDCF against 1 DCF station. |
|
|
|
|
||||
AC |
12 |
|
|
|
|
|
|
|
|
10 |
|
|
|
|
|
|
|
|
|
per |
|
|
|
|
|
|
|
|
|
8 |
|
|
|
|
|
|
|
|
|
thrp |
|
|
|
|
|
EDCF sim. |
|
||
|
|
|
|
|
|
|
|
||
cummulative |
6 |
|
|
|
|
|
EDCF analyt. |
|
|
4 |
|
|
|
|
|
DCF sim. |
|
||
|
|
|
|
|
|
DCF analyt. |
|
||
|
2 |
|
|
|
|
|
|
|
|
|
0 |
|
|
|
|
|
|
|
|
|
0 |
2 |
4 |
6 |
8 |
10 |
12 |
14 |
16 |
|
|
|
Offered traffic per station [Mbit/s] |
|
|
(c) DCF with extended interframe space
[Mbit/s] |
16 |
PF=1.5, AIFS=DIFS. |
|
|
DCF sim. |
|
|||
|
14 |
1 EDCF against 1 DCF station. |
|
EDCF sim. |
|
||||
|
|
EDCF analyt. |
|||||||
|
|
|
|
|
|
|
|||
AC |
12 |
|
|
|
|
|
DCF analyt. |
|
|
|
|
|
|
|
|
|
|||
|
|
|
|
|
|
|
|
||
per |
10 |
|
|
|
|
|
|
|
|
8 |
|
|
|
|
|
|
|
|
|
thrp |
|
|
|
|
|
|
|
|
|
6 |
|
|
|
|
|
|
|
|
|
cummulative |
|
|
|
|
|
|
|
|
|
4 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
|
|
|
|
|
|
|
|
0 |
|
|
|
|
|
|
|
|
|
0 |
2 |
4 |
6 |
8 |
10 |
12 |
14 |
16 |
|
|
|
Offered traffic per station [Mbit/s] |
|
|
(b) EDCF with smaller PF
(d) backoff delays in saturation (simulation results)
lines w/o markers: analytical results (a-c), lines with markers: WARP2 simulation results
Figure 5.21: Throughput and backoff delay results for one EDCF backoff entity contending with one legacy DCF station. The analytical results give the saturation throughput per AC only, which is here and in the following figures of this section indicated as maximum achievable throughput when the offered traffic is increased.
observes always a smaller backoff delay than the legacy DCF station. It can be further observed that in the last scenario, where the legacy DCF station uses EIFS instead of DIFS, the EDCF backoff entity observes significantly smaller backoff delays.
|
using high |
using DIFS, or |
priority with |
EIFS instead |
|
|
different |
of DIFS |
EDCF |
PF |
|
backoff |
|
|
entity |
|
|
|
|
DCF |
|
|
legacy |
receiving station |
stations |
Figure 5.22: Scenario. All stations detect each other. If two or more stations transmit at the same time, a collision occurs.