Binds channel access to different traffic classes by prioritized contention parameters
Fair scheduling based
Partitions the channel bandwidth fairly by regulating wait times of traffic classes in proportion according to given weights
Tunable parameters :
Contention window (Cwmin/Cwmax), backoff algorithm, interframe space AIFS
Service Differentiation mechanisms - EDCF
Enhanced DCF (EDCF)
Part of upcoming 802.11e standard
Priority of traffic categories based on
AIFS (Arbitrary Interframe Space)
Max/min Contention Window
Multiplication factor for backoff window
Combination of above parameters is permitted according to the service provider needs
Although all traffic categories keep using the same DCF access method they have different probabilities of winning the channel contention by differentiating contention parameters.
Service Differentiation mechanisms –P-DCF
Persistent Factor DCF (P-DCF) .
Each traffic class is associated with a persistent factor P
In backoff stage a uniformly distributed number r is generated.
Each flow stops backoff and starts transmission only if r>P in the current slot time, given no transmission occurs in previous slot times.
Therefore the backoff interval is geometrically distributed random variable with parameter P
Service Differentiation mechanisms DWFQ Algorithm 1
Distributed Wighted Fair Queue (DWFQ) –algo 1
Backoff CW of any traffic flow is adjusted based on the difference between the actual and expected throughputs.
If actual_thru<exp_thru then decrease CW in order to increase flow’s priority and vice versa
Service Differentiation mechanisms DWFQ Algorithm 2
Distributed Wighted Fair Queue (DWFQ) –algo 2
Calculation of Li=Ri/Wi
Ri actual throughput, Wi corresponding weight of I station
Comparison of Li with all others and adjust CW for station I, ie if Li<all others decrease CW of I
Service Differentiation mechanisms -DFS
Distributed Fair Scheduling (DFS)
Differentiate the backoff interval (BI) based on the packet length and traffic class and station with smaller BI transmits
BIi= Ρ i x scaling x factor x Li/ Φ i
Li packet length, Φ i weight, Pi random variable uniformly distributed in [0.9,1.1]
Pi is introduced to minimize the collision caused by multiple stations with the same BI
Service Differentiation mechanisms -DDRR
Distributed Deficit Round Robin (DDRR)
The i throughput class at the j station is assigned with a quantum rate ( Qi,j ) equal to the thoughput it requires and a deficit counter ( DCi,j ) that accumulated at the rate of Qi,j and is decreased by the packet length whenever a packet is transmitted.
DCi,j is used for Interframe Space ( IFSi,j ) calculation
IFSi,j is the waiting time before transmission or backoff starting
A larger DCi,j results in a smaller IFSi,j
Service Differentiation mechanisms – Conclusion and Comparison
Fair scheduling based
Fairly allocation of bandwidth among traffic classes.
Prevent starvation of specific class.
Often require a substantial modification of existing 802.11 standards.
Require less modification of the existing DCF access method.
Provide better QoS support for real time applications.
Service differentiation does not perform well under high traffic loads due to the inefficiency of 802.11 MAC.
QoS mechanisms for Admission Control and Bandwidth Reservation (MAC)
Necessity in order to guarantee QoS in high traffic load.
A wireless node has no knowledge of the exact network condition.
With contention based CSMA/CA, bandwidth provisioning is almost impossible only soft QoS guarantee.
In general admission control requires less modification than bandwidth reservation in 802.11 standards.
Admission Control and Bandwidth Reservation
Admission Control Approaches
Measurement-based (admission control based on measurements of existing network status)
Virtual MAC [Barry et al]: Channel is passively monitored by virtual MAC frames and local service (throughput/delay) is estimated by the measurement of virtual frames.
Probe Packets [Valaee & Li]: Admission procedure is based on a sequence of probe packets for ad hoc networks
Data Probe Packets:[Shah et al]: Data packets for measuring the network load
Calculation based approaches
Calculation-based (there are certain perfomance metrics or criteria for evaluating network status)
Permissible throughput: [Kazantzidis et al]: Admission decision criterion is permissible throughput. AODV routing protocol
Saturation-based[Zhu et al]: Prediction and prevention of saturation using piggybacked information (number of active stations, corresponding bit rates, average packet lengths) for each station. Admission control decisions are made dynamically at both source/destination in a fully distributed way.
Flow reservation and priority allocation [Li and Prabhakaran]: Optimizing the usage of priority resources.
ARME [Banchs & Perez]: Based on an extension of DCF. A Token-bucket algorithm is used to detect overload and improve performance via adjustment of CW.
AACA[Liu et al]: ACCA adopted the RTS/CTS access method on a common channel solely for reservation purposes. After successful reservation a pair station transmit without interruption in the reserved channel.
QoS Mechanisms for Link Adaptation (PHY)
802.11 specifies multiple transmission rates that are achieved by different modulation techniques in the PLCP header of PHY layer.
Rate adaptation and signaling are open.
Link adaptation mechanism has to maximize the throughput under dynamically changing channel conditions.
So an obvious solution is to focus on switching transmission rates specified at PLCP, without modification of existing standards.
QoS Mechanisms for Link Adaptation (PHY) –cont’d
Adjust the length of DSSS pseudo noise (PN) in 802.11b with slight modifications of 802.11b
Metrics used in Link Adaptation (PHY)
Channel signal to noise ratio / carrier to interference ratio (SNR/CIR).
Received power level.
Average payload length
Combinations of above
Received Signal Strength RSS Pavon and Choi
Transmission power is fixed.
Linear relationship between average RSS and SNR.
A rate adaptation algorithm at every station maintain its own 12 RSS thresholds and corresponding rates. Based on the measured RSS a station dynamically switches to an appropriate transmission rate.
MPDU based Qiao et al
A combination of metrics is used including:
Average payload length
Frame retry count
The proposed algorithm pre-established a table of best transmission rate for decision making.
Success/Fail thresholds Cheville et al
Transmitted frames ACKs are used as a metric of channel condition.
If number of consecutive successful exceeds S, transmission rate is increased, otherwise transmission rate is decreased.
ACKs are used to indicate transmission success or fail.
Code Adapts To Enhance Reliability Mullin et al
CATER adaptive PN algorithm gives a throughput improvement under high bit error rate (BER) channel conditions at 802.11b.
But due to signaling overhead the throughput under low BER channel conditions is lower than the standard 802.11b
Challenges and future work Wireless Internet and Interoperability
IEEE 802.11 WLANs have been successfully applied as the last mile technology where there is a need for wireless/mobile users (Wireless Hotspots).
There is an urgent demand for e2e QoS guarantee to be provided in wire-cum-wireless heterogeneous networks.
Interoperability between IEEE 802.11 and DiffServ or IntServ.
Wireless Internet and Interoperability
802.11 and DiffServ [Park & Kim]
A proposed architecture for e2e QoS across wired WAN, wired LAN and WLAN.
802.3 MAC frame carry the user priority via the 802.1Q VLAN tag. User priority is forwarded through 802.1D MAC bridge to 802.1e and used by EDCF to differentiate flows.
802.11 and DiffServ [Park & Kim] cont’d
There is a necessity for mapping between DiffServ Code Point (DSCP) and Traffic Category Identification (TCID) defined in 802.1e.
When IP packets are encapsulated in MAC frames they are placed in priority queues without preemption
IP packets are classified and shaped according to the priority of DSCP values before being forwarded to 802.1e priority queues more accurate e2e QoS .
802.11 and IntServ [Liu & Zhu]
Integration of RSVP and WRESV (WRESV is a proposed MAC layer flow reservation and admission control protocol in IEEE 802.11WLAN).
Message mapping at Access Point are implemented by cross-layer interaction and user priorities are mapped to 802.11 MAC with 802.1p.
WRESV is working with most of the existing MAC schedulers (DCF, EDCF, DFS).
This scheme also considers support of both node mobility and QoS in handoff.
Support of Full Mobility
Mobility is supported through extended service set (ESS) for roaming among multiple AP. This roaming capability is achieved through Mobile Station (MS) beacon scanning in a channel sweep.
802.11 WLAN service is only available for low mobility devices in isolated hot spots.
Recent efforts have been made to extend 802.11 WLANs into outdoor cellular networks to provide fully mobile broadband service with ubiquitous coverage and high speed connectivity.
Support of Full Mobility - Examples
[Leung et al] claim that without standard modification the DCF access method with RTS/CTS is feasible for large outdoor cellular coverage (service area 6 Km).
Beam transmission instead of in all directions extends the coverage of 802.11 [Vivato Inc]. Cisco Aironet, Motorola Inc Canopy Radio and Proxim may reach up to 10 Km.
But a cell with large outdoor coverage does not guarantee high speed connectivity due to unavoidable channel contention throughput may degrade in overcrowded cell.
QoS and Mobility Management in Hybrid Wireless Networks
Seamless horizontal handoff and roaming among 802.11 WLAN supporting QoS anytime anywhere.
Vertical handoff between WLAN, mobile and ad hoc networks (MANET), Bluetooth, Universal Mobile Telecommunications System (UMTS) and Wideband Code Division Multiple Access (WCDMA)
WLAN – MANET Integration
[Lamont & Wang]:
Routing within MANET is handled by the Optimised Link State Protocol (OLSP).
Handoff between MANETs & WLANs is supported through automatic node detection and node switching capabilities of the mobiles.
Functionalities of OLSP are extended to support Mobile IPv6
WLAN – Bluetooth Integration
[Conti & Dardari]:
Analytical model for evaluation of the interference between IEEE 802.11 & Bluetooth.
In the proposed model PHY and MAC layers are considered, and the model can be easily implemented.
Performance is evaluated by packet error probability in terms of the relative distanced between the two systems for different conditions.
WLAN – 3G Integration
A mobile node is maintaining two connections in parallel :
Data connection through WLAN.
Voice connection through UMTS.
Vertical handoff between WLANs and CDMA
Real time traffic takes into account handoff delay
Best effort traffic takes into account throughput.
WLAN – 3G Integration cont’d
[Buddhicot & Chandranmenon]:
Combination of the features of high rate small-coverage WLAN and wide-coverage low rate 3G to improve the QoS and flexibility of wireless service.
A loose integration approach is realized with an IOTA gateway and a new client software in order to support seamless mobility, OoS guarantees and multiprovider roaming agreements.
WLAN – 3G Integration IOTA gateway
WLAN – 3G Integration cont’d
Integration of WLANs and 3G/4G requires a low call dropping probability in the 3G/4G networks.
[Lou & Li]:
Adaptive allocation scheme termed measurement based preassignment in order to prevent handoff failure in wireless cellular networks. A periodic measurement of traffic status within a cell help to adjust the number of reserved channels for handoff.
Classification of QoS schemes
Link adaptation in the PHY layer.
Channel access coordination in the MAC layer.
Admission control strategies in MAC & higher layers.