BASED SIMULATION STUDY OF TCP FAIRNESS IN MULTI-HOP WIRELESS NETWORKS
Kunal Shah Advisor: Dr. Harish Sethu
Computer Communications Laboratory
Click to edit Master title Outline style
Introduction and Motivation Click to edit Master text styles Background Second level Model Description Third level Simulation Results and Analysis Fourth level Conclusion Fifth level
6/6/03
2
Click to edit Master title Introduction style
Ad Hoc Networks: Click to edit Master
Second level as a host and a router simultaneously Capable of acting Communicate with each other over shared, multi-hop wireless channels Third level a fixed wired or wireless infrastructure is Required where either unavailable or destroyed Fourth levelby high mobility, low bandwidths, limited Characterized physical security Fifth level and continuously changing network topology
A collection of nodes
text styles
Once the ad hoc network is up and running using some routing protocol, next step is to evaluate performance of transport layer protocol
6/6/03
3
Click to edit Master title Motivation style
As local area wireless networks based on Click to edit Master text styles IEEE 802.11 standard see increasing public Second levelis important to ensure that deployment, it accesslevel Third to network by different users remains fair Fourth level studies devoted to formal No structured investigation of TCP fairness in wireless Fifth level multi-hop networks
6/6/03
4
Click to edit Master title Focus style
Evaluate edit Master Reno, New Reno and Click to TCP Tahoe, text styles SACK for fairness
Analyze the effects of packet size, load, TCP receive buffer size and RTS/CTS on TCP fairness
6/6/03
Motivation for selecting Second level these TCP implementations was their popularity Fairness metric based on maximal Third level normalized distance between users ideal Fourth level share of service delivered share and actual by network Fifth level
5
Click to edit Master title TCP style Evolution
Dominant reliable transport styles since Click to edit Master text protocol its origin Second level Consists of sliding window mechanism, Third level which, in conjunction with ACKs and sequence numbers, guaranteed a reliable Fourth level delivery and flow control Fifth level No congestion control or avoidance mechanism
6/6/03
6
Click to edit Master title TCP Evolution (Cont.) style
text virtually the Click to edit MasterAIMD styles base of all existing Second level TCP protocols Third level Besides maximizing link bandwidth, TCP Fourth level must be fair to rest of Fifth level the flows
Efficient TCP is not guaranteed to be fair
6/6/03
7
Click to edit Master title TCP Tahoe style
Congestion Master text styles Click to edit Control level Second Algorithms: Third Start Slow level Congestion Fourth level Avoidance Fast Retransmit Fifth level Problem:
Transits to slow start after each packet loss
6/6/03
8
Click to edit Master title TCP Reno style
Extension of TCP Tahoe styles Click to edit Master text
Second level Third level Fourth level Fifth level
Added Fast Recovery along with Fast Retransmit
TO: Time-out TD: Threshold Duplicate
6/6/03
9
Click to edit Master title TCP New Reno style
TCP Reno problem: text styles Click to edit Master
Second level TCP New Reno remains in fast recovery Third level mode despite receiving partial Fourth level acknowledgement after fast retransmission Retransmits Fifth levelat the rate of one packet per RTT until all the lost packets are retransmitted
No retransmit timeout
Fast Recovery algorithm rendered inefficient in the presence of multiple losses within a single transmission window
6/6/03
10
Click to edit Master title TCP SACK style
Selectiveedit Master text stylesused to Click to Acknowledgements are provide the sender with sufficient Second level information to recover from multiple packet Third level a single transmission window losses within
Sender knows exactly which Fourth to quickly recover packets to retransmit and so level is able Fifth level Problem: Inefficient in the case of small sender window size
6/6/03
11
Click to edit Master title Fairness Criteria style
Intuitively, oneMaster text fairness as the Click to edit can think of styles closeness of achieved throughput to its fair Second level share
Third level Fourth level Fifth level
6/6/03
12
Click to edit Master title Max-Min Fairness (MMF) style
When flows have equal styles Click to edit Master text weights, MaxMin Fair level allocation can be Second share defined as: Third level allocated in order of increasing demand Resources are No user gets a Fourth levelresource larger than its demand Users with unmet demands get an equal share of the Fifth level resource
6/6/03
13
Click to edit Master title MMFstyle Example
Dividing a 8 slice pizza styles Click to edit Master textamong 4 people Second level slices 2 2 slices Third level 4 slices 2 slices + 1 slice = 3 slices Fourth level desires and gets Fifth level
3 slices 2 slices + 1 slice = 3 slices 1 slice 4 slices
6/6/03
2 slices - 1 slice = 1 slice
14
Click to edit Master title style Proposed Unfairness Criterion
Click to edit Master text styles Second level = max A F U i F Third level where Fourth level = MMF (C, d , d , , d ) F Fifth level
i i i
i i 1 2 n
6/6/03
15
Click to edit Master title style Sample Unfairness Calculation
Click to edit Master text styles 2 level 0.8 Mbps 0.6 Mbps SecondMbps Third level Fourth3level 0.7 Mbps Mbps 0.6 Mbps Fifth level
1 Mbps 0.3 Mbps
6/6/03
i
di
Ai
MMFi
C = 1.8 Mbps
U = 0.5
0.6 Mbps
16
Click to edit Master title style Related work on TCP Fairness
When flows with different end-to-end propagation delays shared a link, the bandwidth allocation was far from being fair
Click to edit Master text styles Constant rate window increase algorithm Increase-by-K policy SecondAvoidance with Normalized Interval of Time (CANIT) level Congestion Wireless links are characterized by long RTT and above Third level opening up the congestion window at a schemes react by much higher rate Increased probing harmful as slow Fourththemselves a bottleneck in56k modem links and band-limited wireless level links are the network Performance degradation not only due to transmission errors and losses but also due congestion Fifthtolevel at base station Fast TCP amount of bandwidth Flow that got head-start occupied large
and starved the flows starting later on split buffer queues Unfair packet dropping policy at Internet routers RED policy
6/6/03
17
Click to edit Master title Related work style Performance on TCP
Click to edit Master text styles Explicit Congestion Notification (ECN) Second level Explicit Link Loss Notification Third level M-TCP level Fourth Split TCP Fifth level
Snoop TCP
6/6/03
18
Click to edit Master title MAC Layer Fairness style
IEEE 802.11 uses per-node queue with per node back-off edit Master text styles Click to
Second level Third level Fourth level Implementing changes made to MAC layer are impractical given the wide deployment of wireless Fifth level networks using 802.11 IEEE standard
Head-of-line packet headed towards a receiver that is in high contention neighborhood can block other flow transmissions to lightly loaded neighbors Node with many flows penalizes its flow unfairly Flows that experience more contention will block more contending flows while transmitting
Lot of research done in improving MAC layer fairness and TCP performance but no real effort made in studying TCP fairness
6/6/03
19
Click to edit Master title Model Description - Node style
Click to edit Master text styles Second level Third level Fourth level Fifth level
6/6/03
20
Click to edit Master title Applicationstyle Process Model
Click to edit Master text styles Second level Third level Fourth level Fifth level
6/6/03
21
Click to edit Master title TCP Process Model style
Click to edit Master text styles Second level Third level Fourth level Fifth level
6/6/03
22
Click to edit Master title AODV Background style
Source-basedMaster protocol based on Click to edit routing text styles DSDV and DSR
Second level number of DSDV and on-demand route Utilizes sequence discovery and Third level maintenance mechanisms of DSR Power Efficient Fourth levelperiodic update messages No flooding or Fifth level
6/6/03
23
Click to edit Master title AODV Process Model style
Click to edit Master text styles Second level Third level Fourth level Fifth level
6/6/03
24
Click to edit Master title WLAN Process Model style
Click to edit Master text styles Second level Third level Fourth level Fifth level
6/6/03
25
Click to edit Master title Simulation Scenario style
Click to edit Master text styles Second level Third level Fourth level Fifth level
6/6/03
26
Click to edit Master title Simulation Setup style
Every to edit Master text styles Click TCP connection is of type FTP and all flows start at the same time Second level if user 1 (node 1) wants to In each scenario, send x Mbps, then user 2 wants to send 2x Third level3 wants to send 3x Mbps and Mbps, user so forth level Fourth Mobility pattern is static Fifth level is infinity and transmitter Battery power power is 0.25 Watts
6/6/03
27
Click to edit Master title Simulation Setup (Cont.) style
Packet sizes are varied from 128 bytes to Click to edit Master text styles 1,024 bytes but ACKs are kept at 40 bytes long Second level TCP receive buffer size is varied from 8,760 bytes to 131,072 bytes Third level Load is varied from 1.5 Mbps to 7.5 Mbps to Fourth level simulate low, medium and high traffic loads Fifth level Load is varied from 1.5 Mbps to 7.5 Mbps but with RTS/CTS enabled for packet sizes larger than 255 packets
6/6/03
28
Click to edit Master title Simulation Setup (Cont.) style
All other parameters were left unchanged as Click to edit Master text styles per IEEE 802.11b standard Second level Simulation was conducted for TCP Tahoe, Reno, levelReno and SACK Third New
Fourth level Fifth level
6/6/03
29
Click to edit Master title Simulation Resultsstyle Receive Buffer Size TCP
Click to edit Master text styles Second level Third level Fourth level Fifth level
Unfairness Distribution for TCP Tahoe
2 1.5 1 Unfairness 0.5 0 -0.5 -1 -1.5 8760 32768 65535 131072 TCP Receive Buffer Size (bytes)
user1 user2 user3 user4 user5
6/6/03
30
Click to edit Master title Simulation Resultsstyle Receive Buffer Size TCP
Click to edit Master text styles Second level Third level Fourth level Fifth level
Unfairness Distribution for TCP SACK
2 1.5 1 Unfairness 0.5 0 -0.5 -1 -1.5 8760 32768 65535 131072 TCP Receive Buffer Size (bytes)
user1 user2 user3 user4 user5
6/6/03
31
Click to edit Master title style Simulation Results Load with No RTS/CTS
Click to edit Master text styles Second level Third level Fourth level Fifth level
Unfairness Distribution for TCP Tahoe
2.5 2 1.5 1 Unfairness 0.5 0 -0.5 -1 -1.5 1.5 4.5 Load (Mbps) 7.5
user1 user2 user3 user4 user5
6/6/03
32
Click to edit Master title style Simulation Results Load with RTS/CTS
Click to edit Master text styles Second level Third level Fourth level Fifth level
Unfairness Distribution for TCP Tahoe with RTS/CTS
1.5 1 0.5 Unfairness 0 -0.5 -1 1.5 4.5 Load (Mbps) 7.5
user1 user2 user3 user4 user5
6/6/03
33
Click to edit Master title sty...