5-1

CS – 647: Advanced Topics in Wireless Networks

Drs. Baruch Awerbuch and Amitabh Mishra

Computer Science DepartmentJohns Hopkins

Transport Layer for Mobile Ad hoc Networks

5-2

Reading Chapter 7 – Ad Hoc & Sensor

Networking, Cordeiro & Agrawal, 2007 One of the suggested text for the course

5-3

Outline Overview of TCP The problems of TCP over MANETs Overview of best transport protocols In depth

Specific problems of TCP over MANETs Details of major TCP variants Discussion - other efforts

Conclusion

5-4

TCP in Wired Network and MANET

Data stream in Wired Network

ACKs stream

Data stream in a MANET

ACKs stream

TCP Source

TCP Sink1 N

432

5-5

Network Architecture at a Crossroads Wireline-centric network design is

“obsolete” New network environments have

emerged Ad hoc, sensors, consumer-owned, delay-tolerant

New networking technologies have emerged UWB, cooperative approaches, MIMO, directed

antennas

The R&D community recognizes the need for change

Introduction

5-6

Revisiting the Current Transport Architecture The vision:

Wireless as an integral part of the network Multiple wireless hops: not just the last mile

(Cellular) Pockets of wireless ad hoc connectivity

A new protocol stack is required Is TCP/IP capable of delivering?

Introduction

5-7

Problem Statement Why does TCP perform poorly in

MANETs? Developed for Wireline networks Assumes all losses congestion related

Many TCP variants have been proposed How good are they? Are they sufficient?

Are there any other alternatives? Are non-tcp protocols the solution?

5-8

Our Goal Identify the problems of TCP in MANETs. Evaluate various major TCP variants.

12 TCP variants, 7 improvement techniques Observations:

Most TCP variants are NOT sufficient. A new transport layer protocol may be/is

needed.

5-9

TCP Basics Byte Stream Delivery

Connection-Oriented: Two communicating TCP entities (the sender and the receiver) must first agree upon the willingness to communicate

Full-Duplex: TCP almost always operates in full-duplex mode, TCP exhibit asymmetric behavior only during

connection start and close sequences (i.e., data transfer in the forward direction but not in the reverse, or vice versa)

5-10

Reliable TCP Guarantees A number of mechanisms help provide the guarantees:

Checksums: To detect errors with either the TCP header or data

Duplicate data detection: Discard duplicate copies of data that has already been received

Retransmissions: • For lost and damaged data • Due to lack of positive acknowledgements• Timeout period calls for a retransmission

Sequencing: To deliver the byte stream data to an application in order

Timers: Various static and dynamic timers used for deciding when to retransmit

Window: For flow control in the form of a data transmission window size

5-11

Conventional TCP: Tahoe, Reno, New-Reno

Sending rate is controlled by Congestion window (cwnd): limits

the # of packets in flight Slow-start threshold (ssthresh):

when CA start Loss detection

3 duplicate ACKs (faster, more efficient)

Retransmission timer expires (slower, less efficient)

Overview of congestion control mechanisms Slow-start phase: cwnd start from 1

and increase exponentially Congestion avoidance (CA): increase

linearly Fast retransmit and fast recovery:

Trigger by 3 duplicate ACKs

Congestiondetected

Fast retransmit/fast recovery

1

2

3

4

threshold

threshold

Time

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 220

2

4

6

8

10

12

14

16

18

20

22

24

26

28

30

32

34

Con

gesti

on

win

dow

s s

ize

Overview

Slow-start Congestionavoidance

Overview of TCP Concepts

5-12

TCP BasicsCongestiondetected

Fast retransmit/fast recovery

1

2

3

4

threshold

threshold

Time

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 220

2

4

6

8

10

12

14

16

18

20

22

24

26

28

30

32

34

Con

ges

tion

win

dow

s si

ze

Slow-start Congestionavoidance

5-13

Congestion Control Slow Start (SS): A mechanism to control the

transmission rate) When TCP connection starts (Initial Value): CWND

=1, congestion window increases by one segment for

each acknowledgement returned

Congestion Avoidance(CA): Used to reduce the transmission rate When Slow Start drops one or more packets due to

congestion

Fast Retransmit: Sender receiving triple duplicate ACKs Immediate transmission of missing packet without

waiting for the Retransmission Timeout to expire

Fast Recovery: In SS or CA when sender receiving triple duplicate ACKs Sender only enters Congestion Avoidance mode

5-14

1. Mobility Route stability and availability

2. High bit error rate Packets can be lost due to “noise”

3. Unpredictability/Variability Difficult to estimate time-out, RTT, bandwidth

4. Contention: packets compete for airtime Intra-flow and inter-flow contentions

5. Long connections have poor performance More than 4 hops thruput drops dramatically

Overview

What is Different in MANETs?

5-15

TCP-Westwood [Casetti et. al.]

Estimate bandwidth to alleviate the effect of wireless errors. TCP-Jersey [Xu et. al.]

Estimate bandwidth to alleviate the effect of wireless errors. Congestion warning assists the determination of packet loss

due to wireless error from congestion. ATP [Sundaresan et. al.]

Rate based transmission, periodic rate feedback, no timeout concept, reliability provided by SACK.

Split-TCP [Kopparty et. al.]

Separating congestion control from reliability. Dropped packets are recovered from the most recent proxy

instead of the source.

Overview

Overview of Best Protocols

5-16

Why Does TCP Fail in MANETs?Specific problems are identified:1. TCP misinterprets route failures as congestion2. TCP misinterprets wireless errors as congestion

3. Intra-flow and inter-flow contention reduce throughput and fairness

4. Delay spike causes TCP to invoke unnecessary retransmissions

RTO too small unnecessary retransmissions.

5. Inefficiency due to the loss of retransmitted packet When retransmitted packet is lost timer expires

performance drops

5-17

Specific problems of TCP over MANETs

TCP misinterprets route failures as congestion Effects: Reduce sending rate Buffered packets (Data and ACKs) at

intermediate nodes are dropped. Sender encounters timeout.

• Under prolonged disconnection, a series of timeouts may be encountered.

TCP in MANET

5-18

TCP misinterprets wireless errors as congestion Effects: Incorrect execution of congestion

control Performance drops. Wireless channel is error-prone compared to

wireline• Fading, interference, noise

Specific problems of TCP over MANETs

TCP in MANET

5-19

Intra-flow and inter-flow contention Effects: Increased delay, unpredictability, and

unfairness. Inter-flow contention: contention of nearby flows. Intra-flow contention: between packets of the same

flow (e.g. forward data and reverse ACKs). Wireline: only packet on same link “compete” Wireless: all close by devices compete for the channel

Data stream

ACKs stream

Specific problems of TCP over MANETs

Two nearby flows

TCP in MANET

5-20

Drawback of TCP Exponential Back Off

5-21

No communication between the partitions

Impact of Partition on Throughput

SA

B CP

X YZ

D

Link Failure

Data transfer continues in spite of failure

5-22

Effects of Partitions on TCP

Node 5 moves away from node 3 (short-term partition)

5-23

1

2

3

4

5

6

7

8

9

The routing protocol reestablishes the path through node 6

Reestablishing Path

5-24

Node 5 moves away from node 3 (long-term partition)

Long Term Partition

5-25

No communication between the partitions

Long Term Network Partition

5-26

TCP Throughput

TCP throughput is inversely proportional to the number of hops

Larger the number of nodes a TCP connection needs to span, lower is the end-to-end throughput, as there will be more medium contention taking place in several regions of the network

5-27

Impact of Lower Layers on TCP -MAC It is intended for providing an efficient shared

broadcast channel through which the involved mobile nodes can communicate

In IEEE 802.11, RTS/CTS handshake is only employed when the DATA packet size exceeds some predefined threshold

Each of these frames carries the remaining duration of time for the transmission completion, so that other nodes in the vicinity can hear it and postpone their transmissions

The nodes must await an IFS interval and then contend for the medium again

The contention is carried out by means of a binary exponential backoff mechanism which imposes a further random interval

At every unsuccessful attempt, this random interval tends to become higher

5-28

Consider a linear topology in which each node can only communicate with its adjacent neighbors

In addition, consider that in Figures (a) and (b) there exist a single TCP connection running between nodes 1 and 5

Impact of Lower Layers on TCP -MAC

5-29

Capture Conditions

In Figure (c) where there are two independent connections,(connection 2-3) (connection 4-5)

Assuming that connection 2-3 experiences collision due to the hidden node problem caused by the active connection 4-5 , node 2 will back off and retransmit the lost frame

At every retransmission, the binary exponential backoff mechanism imposes an increasingly backoff interval, and implicitly, this is actually decreasing the possibility of success for the connection 2-3 to send a packet as connection 4-5 will “dominate” the medium access once it has lower backoff value

In consequence, the connection 2-3 will hardly obtain access to the medium while connection 4-5 will capture it

5-30

Network Layer Impact

Routing strategies play a key role on TCP performance

There have been a lot of proposed routing schemes and, typically, each of them have different effects on the TCP performance

5-31

DSR DSR protocol operates on an on-demand basis in

which a node wishing to find a new route broadcasts a RREQ packet

The problem with this approach concerns the high probability of stale routes in environments where high mobility as well as medium constraints may be normally present

The problem is exacerbated by the fact that other nodes can overhear the invalid route reply and populate their buffers with stale route information

It can be mitigated by either manipulating TCP to tolerate such a delay or by making the delay shorter so that the TCP can deal with them smoothly

5-32

Path Asymmetry Impact In Ad hoc networks, there are several

asymmetries

Loss Rate Asymmetry: It takes place when the backward path is significantly more error prone than the forward path

Bandwidth Asymmetry: Arises when forward and backward data follow distinct paths with different speeds Can happen in ad hoc networks when all nodes not

have the same interface speed

Media Access Asymmetry: Arises when TCP ACKs and Data are contending for the same

5-33

Route Asymmetry Route asymmetry implies having different paths in

both directions

Route asymmetry is associated with the possibility of different transmission ranges for the nodes

The inconvenience with different transmission ranges is that it can lead to conditions in which the forward data follow a considerably shorter path than the backward data (TCP ACK) or vice versa --> affecting hop counts and delays (RTT)

Multi-hop paths are prone to have lower throughput and TCP ACKs may face considerable disruptions

5-34

Overview of Results The best TCP variants:

TCP-Westwood and TCP-Jersey seem the best. Both protocols estimate bandwidth more accurately.

TCP mechanisms: Feedback from intermediate nodes leads to big gains.

The best non-TCP approaches: Ad-hoc Transport Protocol (ATP) seems to address

most issues• Non-window based: estimates achievable rate periodically

Split-TCP: promising new way of looking at transport layer

• Dynamically buffer packets mid-path Key: Separation of congestion control from reliability.

5-35

TORA TORA has been designed to be highly dynamic by

establishing routes quickly and concentrating control messages within a small set of nodes close to the place where the topological change has occurred

TORA makes use of directed acyclic graphs, where every node has a path to a given destination and established initially

This protocol can also suffer from stale route problem similar to the DSR protocol

The problem occurs mainly because TORA does not prioritize shorter paths, which can yield considerable amount of out-of-sequence packets for the TCP receiver, triggering retransmission of packets