Wireless Local Area Networks (Part 2)
Reliable Data Delivery
NES440 Wireless Networks
Dr. Fahed Awad
Department of Network Engineering & Security
Jordan University of Science and Technology
Reliable Data Delivery
The IEEE 802.11 physical and MAC layers are subject to data transfer unreliability due to:
Noise, interference, and/or other propagation-related effects that may result in loss of frames (i.e.; drop of an erroneous frame)
Even with relatively strong error-correction techniques, frames may not be successfully received (i.e.; error-free)
This can be dealt with at a higher layer such as the TCP. However:
The retransmission timers at the higher layers are typically in the order of seconds (i.e.; relatively long retransmission delay)
The TCP reliable data transfer was designed to overcome data loss caused by traffic congestion not due to collisions
Therefore, it is more efficient to deal with such errors at the MAC level
The IEEE 802.11 MAC includes a two-way frame exchange protocol:
When a station receives a frame, it returns an acknowledgment (ACK) frame
The exchange process is given a high priority and hence cannot be interrupted by any other station
If no ACK is received within a short period of time, the station retransmits the frame
Four-Frame Exchange
Basic data transfer involves the exchange of two frames (i.e.; the data frame and the ACK frame)
To further enhance the reliability of data transfer, a four-frame (or a four- way frame) exchange may be used such that:
The source STA sends a Request to Send (RTS) frame to the destination STA
The destination STA responds with Clear to Send (CTS) frame
After receiving CTS, the source STA transmits the data
Upon receiving the data, the destination STA responds with an ACK
The RTS alerts all STAs within the range of the source STA that an exchange of frames is about to start
The CTS alerts all STAs within the range of the destination STA that an exchange of frames is about to start
Therefore, the STAs refrain from any transmission for the duration of the frame exchange in order to avoid collisions
RTS/CTS exchange is a required function of the MAC sublayer but it may
be disabled or restricted due its associated large overhead
Wireless Media Access
Wireless Media Access Control
Sharing a media access among many transmitting stations in a wireless network is more complex than a wired network. Why?
Wired stations can detect collisions by sensing the carrier energy on the line and abort transmission accordingly
Therefore, CSMA/CD (Carrier Sense Multiple Access with Collision Detection) is used in wired MAC’s
Wireless stations cannot detect a collision between its transmission and another station’s transmission since the radio cannot transmit and
receive simultaneously
Therefore, CSMA/CA (Carrier Sense Multiple Access with Collision Avoidance) is used in wireless MAC’s
The IEEE 802.11 standard defines two MAC sublayer media access coordination protocols:
The mandatory DCF (Distributed Coordination Function) protocol
The optional PCF (Point Coordination Function) protocol
The 802.11 MAC Coordination Functions
The lower MAC sublayer is the distributed coordination function (DCF)
It is a decentralized MAC algorithm
It is contention-based to provide access to all types of traffic from all stations
It provides access for asynchronous traffic
The optional upper MAC sublayer is the point coordination function (PCF)
It is a centralized polling-based MAC algorithm
It is a contention-free algorithm that provides access to time-critical type of traffic
It is built on top of the DCF
802.11b
IEEE 802.2
The Interframe Spacing
The IEEE 802.11 standard defines standard spacing intervals (or time gaps) between the transmissions of different MAC frames called the interframe spacing (IFS)
There are three types of IFS defined for prioritizing media access:
Short IFS (SIFS)
The shortest IFS
Used for immediate response actions such as:
Acknowledgment (ACK)
Clear to send (CTS)
Poll response
Point coordination function IFS (PIFS)
Has a medium length IFS
Used by the centralized controller in the PCF scheme when using polls
Takes precedence over the normal contention traffic
Distributed coordination function IFS (DIFS)
The longest IFS
The standard interval between transmission of data frames
Used as the minimum delay of asynchronous frames contending for channel access
The Medium Access Control Logic
A station with a frame to send should sense the medium status:
1. If it is idle, wait to see if it remains idle for one IFS. If so, may transmit immediately
2. If it is busy (either initially or becomes busy during IFS), the station should defer transmission and continue to monitor it until the current transmission is over
3. Once the current transmission over, the station should wait another IFS 4. Setup the backoff counter with a random number and keep sensing the
medium
5. If the medium remains idle, decrement the backoff counter every slot time. When the backoff counter reaches zero, the station may start transmitting its frame.
6. During the backoff process, if the medium becomes busy, the backoff timer is halted and it resumes only if the medium becomes idle again for an IFS duration
7. To ensure stability and adaptation to the number of contending stations, binary exponential backoff is used such that if the
transmission attempt fails, the next random number is chosen from a larger set.
8. If there is another frame to transmit (or to retransmit), then wait for an
IFS and jump to step 4
IEEE 802.11 MAC Timing Basic Access Method
The Distributed Coordination Function (DCF)
Wireless networks cannot use CSMA/CD. Why?
For collision detection, the station has to be able transmit and receive at the same time.
This means costly station design
Some collisions may not be detected such as:
Due to the hidden-station problem (to be discussed later)
The distance can be large enough so that the signal fading would prevent the other side from hearing the collision
DCF uses CSMA/CA, a modified version of CSMA/CD, as an access method
CSMA/CA attempts to avoid collisions instead of detecting them but it does not eliminate them
All stations must wait a random amount of time, measured in slot times, after the medium is clear
For example: the IEEE 802.11b slot time is 20μs long. Therefore, if a
station has to wait 4 slot times, then it has to wait 80μs
The Exponential Backoff
It forces the stations to wait for a random amount of time in order to reduce the chance of collision
The backoff period increases exponentially (i.e.; doubles) after each collision, similar to the Ethernet protocol. Why?
If the medium is sensed busy:
Wait for the medium to be idle for a DIFS period
Pick a random number from zero to the contention window (CW) in order to initialize the backoff counter. Initially, CW=1 (but it can be changed by configuration)
Decrement the backoff timer every slot time until it reaches 0
However, whenever the medium is sensed busy, freeze the backoff counter. Why?
When the medium become idle again for at least a DIFS period, resume the counting down.
When the backoff counter reaches 0, transmit the frame
If there is another frame to send, then repeat the process of deferring and contending before transmitting
If the backoff timers of two stations expire at the same slot time, a collision will occur causing the transmission to fail
After every failed transmission attempt:
The CW is doubled
That is, CW = 2i–1, where i is the number of attempts
CW
min= 1 and CW
max= 1023 (e.g., CW = 1, 3, 7,15, 31, …, 1023)
What is the maximum possible waiting time a station may have in IEEE 802.11b?
Answer: 20.5 ms. How?
The DCF CSMA/CA Timing without RTS/CTS
Before transmitting a frame, sense the medium
If it is idle for one DIFS period, then start transmitting
If it is busy, then defer until it becomes idle:
Wait one DIFS period
Go through the random backoff process, then transmit the frame
Wait for the ACK
If it is received within the timeout interval, then you are done with the frame
Otherwise, retransmit after deferring and contending for the channel using the
exponential backoff (remember: after each failed attempt, the CW is doubled)
Example on DCF Timing without RTS/CTS
t busy
bo
estation
1station
2station
3station
4station
5packet arrival at MAC DIFS
bo
ebo
ebo
ebusy
elapsed backoff time bo
rresidual backoff time busy medium not idle (frame, ack etc.)
bo
rbo
rDIFS
bo
ebo
ebo
ebo
rDIFS
busy
busy
DIFS
bo
ebusy
bo
ebo
ebo
rbo
rSending Unicast Packets
The station waits for DIFS before transmitting the data
If the packet is received with correct CRC, the receiver responds with an ACK frame within SIFS
If no ACK is received within SIFS, the data frame is
automatically retransmitted after deferring and contending
t SIFS
DIFS
data ACK
waiting time other
stations receiver
sender data
DIFS
contention
The Hidden Station Problem
Station A is within the transmission range of both stations B and C
Stations B and C are outside the range of each other
If station B is in the middle of sending a frame to Station A:
Station C can’t hear station B’s transmission
Station C has a frame to station A. After a DIFS interval, it will start transmitting the frame to station A causing a collision at station A
In this case, we say that station B and C are hidden from each other with respect to Station A
The hidden station problem increases the probability of packet collision and hence degrades the performance of the WLAN
How is this problem solved?
The RTS/CTS protocol
The RTS/CTS handshake can be optionally used to solve the hidden station problem
Before transmitting, station B sends an RTS frame to station A
Upon receiving the RTS frame, station A responds with a CTS frame
Station C (and any other station within the transmission range of station A) hears the CTS frame and hence refrains from transmitting (and even from sensing the channel) for the whole transmission period
Each station within the range of station A creates a timer called the network allocation vector (NAV) that shows how much time it must wait before it can sense the channel again. This is called Virtual Carrier Sensing
But how does each node know the transmission period?
The CTS/RTS handshake causes a significant overhead on the WLAN especially with short frames
To control this problem, the WLAN may restrict the use of CTS/RTS only to the frames that are
longer than an RTS threshold
The DCF CSMA/CA Timing with RTS/CTS
RTS and CTS frames are smaller than data frames and they use shorter IFS than data frames to guarantee access
Stations that hear either the RTS or the CTS “remember” that the medium will be busy for the duration of the transmission
Based on a Duration field in the RTS and CTS frames
t SIFS
DIFS
data ACK
defer access other
stations receiver
sender data
DIFS
contention RTS
SIFS CTS SIFS
NAV (RTS)
NAV (CTS)
The DCF Timing Example with RTS/CTS
The Exposed Station Problem
It is the opposite of the hidden station problem
A station refrains from using the channel when it is available
Station A is transmitting to station B
Station C hears station A’s transmission and hence refrains from transmitting a frame to D even though it can without causing a collision. How?
The RST/CTS handshake does not solve the exposed station problem
Station C will not be able to hear the CTS frame from station B
Frame Fragmentation
If the channel is noisy (e.g.; a Bluetooth piconet or a microwave oven is close by), corrupted frames have to be retransmitted
The data unit of a large frame is divided into several smaller fragments in order to
Reduce the probability of packet collision
Reduce amount of overall time the medium is in use (i.e.; by reducing the retransmissions)
If the data frame length exceeds a specified threshold, the MAC sublayer fragments it
The receiving station reassembles the fragments
However, if the fragments are too small in size or the impact of noise is not very severe, the fragmentation may degrade the WLAN performance
It is an alternative collision-reduction option to RTS/CTS, but mostly with relatively higher overhead of ACKs and additional SIFS periods
When a station gains access to the channel, it may send all fragments in one burst, where:
Each fragment is individually acknowledged
An SIFS is used between fragments and ACKs
If a fragment is corrupted (i.e.; no ACK is received), the station:
Releases the medium
Contends for the medium again
Starts by retransmitting the unACKed fragments and continues forward
The RTS/CTS handshake may be used to gain access to the medium for transmitting
the fragmented frame
Fragmentation Example 1 with RTS/CTS
t SIFS
DIFS
data ACK
1other stations receiver
sender frag
1DIFS
contention RTS
SIFS CTS SIFS
NAV (RTS)
NAV (CTS)
NAV (frag
1)
NAV (ACK
1) SIFS
ACK
2frag
2SIFS
Fragmentation Example 2 with RTS/CTS
Synchronization in Infrastructure Mode
beacon interval (20ms – 1s)
t medium
access point
busy B
busy busy busy
B B B
value of the timestamp B beacon frame
Synchronization in Ad Hoc Mode
t medium
station
1busy B
1beacon interval
busy busy busy
B
1value of the timestamp B beacon frame
station
2B
2B
2random delay
Point Coordination Function (PCF)
The PCF is an alternative contention-free medium access method that is implemented on top of the DCF in an infrastructure mode
The PCF is used when the wireless network has a number of stations with time-sensitive traffic that need to be given higher priority
The remaining traffic may contend for medium access using CSMA
The PCF is based on polling the stations by a centralized polling master (called a point coordinator or PC), which is usually the AP
The PC polls the intended stations in a round-robin fashion
When a station is polled, it may respond after an SIFS
If the PC receives a response, it issues another poll after an SIFS
The PC uses PIFS (instead of DIFS) to contend for the medium gaining a higher priority than any other station trying to access the medium
Due to the higher priority of the PCF over DCF, contention-based traffic may not have access to the medium during the PCF period
In order to allow contention-based traffic to access the medium, a repetition
interval (also called a superframe) is used, which allows both types of traffic
to access the medium
The Superframe (or Repetition Interval)
At the beginning of the superframe, the PC may seize control and issue polls for a given period, which may vary because of the variable-size frames issued by stations
After seizing control of the medium, the PC starts by sending a beacon frame, which carries information about the duration of the contention-free period (CFP) of the superframe
All stations with contention-based traffic must set their NAV to the CFP
At the end of the CFP, the PC sends a “contention-free end” frame in order to allow the contention-based traffic to use the medium
At end of the superframe interval, the PC may contend for medium access using PIFS
If idle, the PC gains immediate access to the medium and a full superframe period follows
If busy, the PC must wait for the medium to be idle to gain access
This results in a foreshortened superframe period for the next cycle
An Example of a Repetition Interval Timing
ACK is piggybacked on the data or poll frames
Quality of Service and IEEE 802.11e
The DCF does not provide good enough service for real-time or time-sensitive traffic
The demand for multimedia traffic over WLAN is increasing (e.g.; VoIP over WLAN)
Quality of Service (QoS): is the capability to prioritize different types of frames
Wi-Fi Multimedia (WMM) QoS: modeled after a wired network QoS prioritization scheme
The IEEE 802.11e: defines a superset of features intended to provide QoS over WLANs
The IEEE 802.11e proposed two new modes of operation for the 802.11 MAC sublayer
Enhanced Distributed Channel Access (EDCA): Contention-based but for different types of traffic
It has four access categories (ACs)
It provides “relative” QoS but cannot guarantee the service
Hybrid Coordination Function Controlled Channel Access (HCCA):
It is a new form of PCF that is based on polling
It serves as a centralized scheduling mechanism
Power Management
The original IEEE 802.11 standard assumes that the stations are always ready to receive network messages and must remain “awake” to receive network
transmissions
But what if the station is idle? Why does it have to continuously consume energy?
Power management allows mobile devices to conserve battery life without missing network transmissions
It is transparent to all protocols and applications
It differs based on the specific WLAN configuration
AP keeps track of which stations are awake and which are sleeping
Buffering: if a destination station is in sleep mode, the AP temporarily stores the received frames
The station can be configured to one of two power modes:
Continuous Aware Mode (CAM): the station is always awake and is usually used if the station has a continuous power source
Power Save Polling (PSP): the station goes into sleep and awakens with the beacon frames in order to maintain its synchronization with the AP
The periodically broadcasted beacon frame by the AP
Contains the Traffic Indication Map (TIM), which lists all the stations that have frames buffered by the AP
All sleeping stations periodically switch into active listening mode in order to receive the beacon frames