Performance Analysis on 802.11

PERFORMANCE ANALYSIS ON 802.11
Soonmok Kwon
2007-05
1

REFERENCES
 [1] Giuseppe Bianchi, “Performance Analysis of the IEEE 802.11 Distributed Coordination
Function”, IEEE Journal on Selected Areas in Communications, vol. 18, no. 3, 2000
 [2] Omesh Tickoo and Biplab Sikdar, “Queueing Analysis and Delay Mitigation in IEEE
802.11 Random Access MAC based Wireless Networks”, IEEE INFOCOM, 2004
 [3] Hongqiang Zhai, Xiang Chen and Yuguang Fang, “How well can the IEEE 802.11 wireless
LAN support quality of service?”, IEEE Transactions on Wireless Communications, vol. 4, no.
6, pp. 3084-3094, 2005
 [4] H. Zhai and Y. Fang, “Performance of wireless LANs based on IEEE 802.11 MAC
protocols”, in Proc. IEEE Personal Indoor and Mobile Radio Communications (PIMRC), pp.
2586-2590, 2003
 [5] H. Zhai, Y. Kwon, and Y. Fang, “Performance analysis of IEEE 802.11 MAC protocols in
wireless LANs”, Wiley Wireless Commun. Mobile Comput., Special Issue on Emerging
WLAN Technologies and Applications, vol. 4, n. 8, pp. 917-931, December, 2004
2

INDEX
 Introduction
 Throughput
 Normalized throughput
 Saturation throughput
 Bianchi‟s analysis
 Fang‟s analysis
 Delay
 MAC layer service time
 Generalized state transition diagram
 System time
 Notes on 802.11 Behaviors
3

INTRODUCTION
 802.11
 Wireless MAC standard, widely used
 Backoff mechanism is adopted by most of wireless
protocols
 System overview
4
AP
MH
MH MH
MH
 Queueing model
 1 transmission at a time
 Queue: sum of each node‟s queue
 Arrival: sum of data arrival at each node
 Service time: MAC service time
 Hidden terminal
 Original system: occurs
 RTS/CTS: not occurs
n=5

THROUGHPUT – TIME SLOT
 Throughput can be derived with or without Queueing Model
 We use simpler one: without queueing model
 Tx prob. for each node in a slot :
 Here, a slot means
 Idle (Empty) slot time :
 A period with successful tx :
 A period with collision :
 With probability for each:
5
n : node number
starting point

THROUGHPUT - NORMALIZED THROUGHPUT
 Channel performance metrics
 Channel idleness ratio:
 Channel busyness ratio:
 Channel utilization:
 Normalized throughput (goodput):
6

THROUGHPUT – CONDITIONAL COLLISION PROBABILITY
 Conditional collision probability :
 A collision prob. seen by txed packet
 P( collision | tx )
 A cond. coll. prob. with maximum
throughput:
 By plotting, we know the throughput
envelope is convex
 A cond. call. prob. with saturation throughput :
 For specific n, there exist a value of p at which the network operates
in the saturated status.
 This is maximum achievable p for given n.
 Therefore, the cond. coll. prob. with MAX throughput is
7

THROUGHPUT – SATURATION CONCEPT
 Definition of saturation
 We say a network is saturated (with given # of contending nodes)
when the conditional collision probability is maximized.
 If n nodes are maintaining their tx queue non-empty, they will
reach saturation condition.
 For this reason, we sometimes say a node is saturated is it maintains its
tx queue non-empty.
 Traffic load and performance degradation by saturation
 Low load (no inefficiency)
 # of simultaneous access „n‟ is small and MAX(p) < proot
 High load (inefficient due to high contention (p higher than proot))
 # of simultaneous access „n‟ is large and proot < MAX(p)
 Throughput can be enhanced
 See slide 19
8

THROUGHPUT – SATURATION THROUGHPUT (1)
 Bianchi‟s model [1]
9
Empty queue prob.
Tx prob. at saturation
No limit on
retransmission
, if m=0 (no exponential back-off)

 Fang‟s model [3, 4, 5]
10
1
Delay factor:
covered later...

11
Bianchi‟s result
# nodes increase -> MAX(p) increases ->
Throughput decreases after proot
MAC layer service time as a
function of n saturated nodes.
MAX(p) : Coll. prob at saturation t-put

DELAY - OVERVIEW
 MAC layer service time can be
 Modeled with generalized state transition diagram and
 Solved with the Mason formula
 Example
 “Infinite retransmission with any backoff mechanism”
12
Arrival :
Poisson or Deterministic
Service :
MAC layer service time
P(time | success)=Tsuc
P(time | collision) = Tcoll
Mason formula

DELAY – GENERALIZED STATE TRANSITION DIAGRAM (1)
 Successful Tx:
 Tx but collision occurs:
 Decreasing backoff timer by 1:
13
1 * Z^[time which is not related with tx process]
*Z^0

 Successful Tx:
 Tx but collision occurs:
 Decreasing backoff timer by 1:
14
Channel busy among (n-1) stations
Successful Tx occurs among (n-1) stations
Collision occurs among (n-1) stations
Mason
formula

 MAC layer service time in total
15
Mason
formula

DELAY – SYSTEM TIME
 M/G/1
 G/G/1
 ?/G/1
 M/M/1
16

NOTES ON 802.11 MAC BEHAVIORS
Throughput and Delay17

THROUGHPUT - COLLISION PROBABILITY
18Throughput is mostly the function of p
But n controls the maximum achievable p and, thus, controls the
saturation throughput

THROUGHPUT – OPTIMAL AND SATURATED CASE
19Maximum throughput is achieved in the non-saturated case rather
than in the saturated case when n > 5
collision

MAC LAYER SERVICE TIME – COLLISION PROB.
20
Ts directly depends on p. However, n determines maximum p and,
thus, performance at saturated state.
Thus, we can define following function. It‟s very useful!
M(n) : MAC layer service time with given n contenders (saturated nodes)

MAC LAYER SERVICE TIME – APPROXIMATION (1)
21
Lognormal distribution provides a good approximation.
Exponential distribution is reasonably good except with very low p.

22
MAC LAYER SERVICE TIME – APPROXIMATION (2)
Markov chain model over-estimate service time compared to the
simulation results from ns-2. So, the model gives
Lower bound for throughput
Upper bound for delay

CONCLUSION
 WLAN Performance Analysis
 Throughput and Delay are formulated with p, n
 For non-saturated and saturated case
 To understand the throughput, understand the concept of
saturation and conditional collision probability
 To understand the delay, most analysis on CSMA MAC targets
the MAC layer service time with n contenders.
 Using two equations:
 First,
 Second, the equation derived from Markov model for saturation situation
 It determines relationship between tau and p at given protocol
parameters.
23

Performance Analysis on 802.11

More Related Content

What's hot (20)

Viewers also liked (15)

Similar to Performance Analysis on 802.11 (20)

Recently uploaded (20)

Performance Analysis on 802.11