cs6003 ad hoc and sensor networks

CS6003 ADHOC & SENSOR

NETWORKS

UNIT – III

Dr.A.Kathirvel, Professor and Head, Dept of CSE

Anand Institute of Higher Technology, Chennai

Unit - III

ROUTING PROTOCOLS AND

TRANSPORT LAYER IN AD HOC

WIRELESS NETWORKS

Issues in designing a routing and Transport

Layer protocol for Ad hoc networks- proactive

routing, reactive routing (on-demand), hybrid

routing- Classification of Transport Layer

solutions-TCP over Ad hoc wireless Networks.

2

Routing Protocols

for

Ad Hoc Wireless Networks

3

Issues in designing a routing protocol

Mobility

Bandwidth constraint

Error-prone shared broadcast radio channel

Hidden and Exposed terminal problems

Resource constraints

Characteristics of an ideal RP for AWN

4

Characteristics of an ideal RP

It must be fully distributed, as centralized routing involves high control overhead and hence is not scalable. More fault tolerant than centralized.

Frequent topology changes caused by mobility

Minimum connection setup time is desired.

Localized state maintenance

Loop-free and free from stale routes.

No. of packet collisions must be kept to a min

Convergence must be quick

Optimally use scare resources such as BW, power(computing & battery), memory

Provide a certain level of QoS

5

Classifications of Routing Protocols

Routing information update mechanism

use of temporal information for routing

routing topology

Utilization of specific resources

7

Table-Driven Routing Protocols

Extension of wired networks routing protocols

global topology information is maintained in the form

of table at every node

tables are updated frequently in order to maintain

consistent and accurate networks state information.

Example

Destination sequenced Distance-vector RP

Wireless RP

Cluster-Head Gateway Switch RP

Source-Tree Adaptive RP

8

DSDV

Enhanced version of Bellman ford algorithm where each node maintains a table that

contains shortest path from first node to every other node in the networks.

It incorporates table updates with increasing sequence number tags to prevent loops

to counter the counter to infinity problem and for faster convergence

tables are also exchanged frequently to keep an up to date view of the n/w topology

tables are also forwarded if a node observes a significant change in local topology

tables update 1. Incremental update 2. Full dumps Table updates are always initiated

by the destination node with a new sequence number which is always greater than

the previous one.

Upon receiving an updated table, a node either updates it table or holds it for some

time to select the best metric(which may be the lowest no. of hops)

based on the sequence no. of table it may forward or reject table.

Incremental Update Full dumps

It takes a single networks data packet unit

(NDPU)

Multiple NDPUs

Used when a node does not observe a

significant change in local topology

Done either when local topology changes

significantly or when an incremental update

requires more than one NDPU

9

DSDV

10

DSDV The routing table of node1 indicates that the shortest route to the destination node

(node 8) is available through node 4 and the distance to it is 3 hops.

Reconfiguration of path

the end node of broken link initiates a table update message with the broken links

weight assigned to infinity and with a seqno greater than the seqno stored for that

destination.

Each node upon receiving an update with weight, quickly passes it to its neighbors

in order to propagate the broken link information to the whole n/w.

Consider the case when node 5 moves from the current position. When the

neighbor node previous path breaks, it sets all the paths passing thro’ the broken

link with distance as infinity. For ex, when node4 knows about the link break, it

sets the path node 5 as infi and broadcasts its routing table to its neighbors. Those

neighbors detecting significant changes in their routing tables rebroadcast it to

their neighbors. In this way, the broken links information is propagated thro’ the

n/w. when node 8 receives table update mesg from 5, it informs the neighbors

about the shortest distance to node 5. This information also propagated

throughout the n/w. All nodes receiving the new update mesg with the higher

seqno. Set the new distance to node 5 in their corresponding tables.

11

DSDV

Advantages

The availability of routes to all destinations at all times implies

that much less delay is involved in node setup.

Existing wired n/w protocol can be applied to adhoc wireless

n/w with many fever modifications (seqno).

Disadvantages

The updates due to broken links leads to a heavy control

overhead during high mobility. Therefore it is not scalable in

adhoc n/w which have limited BW and whose topologies are

highly dynamic.

In order to obtain information about a particular destination a

node has to wait for a table update mesg initiated by the same

destination node.

12

Wireless Routing Protocol

Similar to DSDV, inherits the prop. Of distributed Bellman ford algorithm

it differs from table maintenance and in the update process

WRP uses a set of tables to maintain more accurate information they are

Distance Table (DT)-distance, predecessor node for a particular destination

Routing Table (RT) - shortest distance, predecessor, successor flag

Link Cost Table (LCT) - cost of relaying through each link

Message Retransmit Table (MRT) - entry for every update msg that is to be retxd

and counter for each entry.

When the link b/n 7 and 9 breaks, all nodes having a route to the D with

predecessor as node 7 delete their corresponding routing entries. Both node 9 & 7

send update msg to their neighbors indicating the cost of the link b/n node 7 &

node 9 is infi. If the nodes have any other alternate route to D 9 they udate their

table and send the changes to its neighbors. A neighbor node after receiving an

update msg, updates its routing table only if the new path is better than the

existing path.

13

WRP

14

WRP

Advantages

WRP has the same advantages as that of DSDV

In addition it has faster convergence and involves fewer table

updates

Disadvantages

The complexity of maintenance of multiple tables demands a

larger memory and greater processing power from nodes.

At high mobility, the control overhead involved in updating

table entries is almost the same as that of DSDV and hence is

not suitable for highly dynamic and also for very large ad hoc

wireless networks.

15

Cluster head Gateway Switch RP

CGSR employs hierarchical networks topology

unlike other table driven routing approaches that uses

flat topologies.

CGSR organizes nodes into cluster

cluster head coordinates all the nodes in cluster

cluster heads are elected dynamically by employing a

Least Cluster Change algo.

A node ceases to be a cluster head only when it

comes under the range of another cluster head, where

the tie is broken either using lowest id or highest

connectivity algo.

16


Different cluster-heads can operate on different

spreading codes on a CDMA systems

Inside a cluster, the cluster head can coordinate the

channel access based on a token-polling protocol

intercluster communication takes place via gateways

the gateways which are members of more than one

cluster can listen to multiple spreading codes.

Every member node maintains a routing table

containing the dest. Cluster head for every node in

the network.

17

CGSR

18


In addition each node maintains a routing table which

keeps track the list of next hop nodes for reaching ever

dest. Cluster.

Each node before sending date gets token from its

cluster head it obtains the dest. Cluster head and the next

hop node from cluster member table and the routing

table respectively.

A path from any node a to any node b will be similar to

a-c1-G1-c2-G2- .. Ci-Gj.. Gn-b where Gi, cj are the ith

gateway and jth cluster head resp.

a path from node 2 to node 10 would follow 2 -1-3-7-10

19

CGSR

Advantages

clustering provides a mech. For allocating the BW.

Hence BW util. Is Better

easy to imp. Priority scheduling schemes with token

scheduling and gateway code scheduling

Disadvantages

Increase in path length and instability in the system at

high mobility when the rate of change of cluster heads

is high

power consumption at the cluster head is high

20

Source-Tree Adaptive Routing Protocol

Proposed by Garcia-Luna-Aceves and Spohn

Variation of table driven rp, with the least overhead routing

approach (LORA) as the key concept rather than the

optimum routing approach (ORA)

ORA - quick update mechanism LORA - Feasible path, not

guaranteed to be optimal, but less overhead.

STAR - Every node broadcasts its source-tree information

source tree of a node consists of the wireless links used by

the node in its preferred path to destinations

Every node, using its adjacent links and the source-tree

broadcast by its neighbors, builds a partial graph of

topology

21

Source-Tree Adaptive Routing Protocol

During initialization, a node sends an update msg to its

neighbors. Also every node is required to originate

update msg about new destination, the chances of routing

loops and the cost of paths exceeding a given threshold.

Hence, each node will have path to every dest node. Path

be sub-optimal

Absence of a reliable link layer broadcast mechanism, it

originates an update msg to all its neighbors indicate the

absence of a path to d. After getting the source tree

update from a neighbor, the node s update its source tree

and, using this it finds a path to all nodes in the network.

22

STAR Presence of reliable broadcast mechanism, STAR -implicit route

maintenance

the link update mech. About the unavailability of a next hop

node triggers an update msg from a neighbor which has an

alternate source tree indicating an alternate next hop node to the

destination.

When an intermediate node receives a Route Repair update msg,

it removes itself from the top of the route repair path and reliabl

sends it to the head of the route repair path.

Advantages

Low overhead among all the table driven routing protocols

use of the LORA approach in this table driven rp reduces the avg

control overhead compared to several other on demand rp

23

On Demand Routing Protocols

Execute path finding process and exchange

routing information only when a path is required

by a node to communicate with destination.

Example

DSR

AODV

TORA

LAR

ABR

SSA

24

Dynamic Source Routing

Eliminates the periodic table-update messages and

thereby reduces the BW consumed by control packets.

Beaconless & hence doesn’t require periodic message

transmission

when a source node has a data packets to be sent to the

dest. It initiates a RREQ

RREQ is flooded throughout the network

each node upon receiving RREQ can fwd it if

it has not fwd the RREQ already

it is not a dest node

25

Dynamic Source Routing Time to live(TTL) of packet has not exceeded

each RREQ carries a seqno generates by S node and

the path it has traversed

a node upon receiving the RREQ checks the seqno

before fwd it.

Seqno is used to avoid loop formations and to prevent

multiple transmission of the same RREQ by

intermediate nodes.

D node after receiving the first RREQ packet; sends a

RREP using the reverse path traversed by the RREQ

packet

26

Dynamic Source Routing

This protocol uses the route cache that stores all

possible info. Extracted from source route contained in

data packet

if an intermediate node receiving a RREQ has a route

to the destination in its route cache it sends RREP with

a complete route from S to D

Optimizations:

1. Route Cache

This cache information is used by intermediate nodes

to reply to the S node when they receive a RREQ and

if they have a route to the corresponding D

27

DSR 2. Promiscuous mode

By operating in this mode, an intermediate node learns abt the path

breaks. Info. Gained is used to update the route cache so that the

active routes maintained in route cache don’t use such links

3. During networks partition

The affected nodes initiate RREQ packets an exponential backoff

algo. Is used to avoid frequent RREQ flooding in the network when

the D is in another dispoint set.

Route maintenance

when an intermediate node moves away causing a wireless link to break. For

ex. If the link between node 5 & 7 fails, a route error msg is generated by a

node adjacent to path break to inform the source node. The source node

reinitiates the route establishment procedure. The cached entries at the

intermediate node and S node are removed when the route error packet is

received.

29

DSR

Advantages

it eliminates periodical table update msg

intermediate nodes utilize the route cache info efficiently to

reduce the ctrl overhead

Disadvantages

route setup delay is more

route maintenance mech doesn’t efficiently repair the path

break efficiently

the performance of this protocol degrades rapidly with

increasing mobility

30

Adhoc Ondemand Distance Vector

AODV uses ondemand approach, ie a route is established

only when it is required by a S node for transmitting data

packet

it differs from DSR from the fact that DSR uses source

routing in which a data packet carries complete path to the D

in AODV, the S node and intermediate nodes stores the next

hop info corresponding to each flow for packet txn

uses dest. Seqno to determine an up-to-date path to the D

a node updates its path info only if the destseqno of the

current packet received is greater than the last destseqnum

stored at the node

31

Adhoc Ondemand Distance Vector

a RREQ carries SID,DID,S-seqno,D-seqno,BcastID and TTL

source 1 initiates the RREQ to be flooded in the nxw for D 15

Assuming that the Dseqno as 3 and Sseqno as 1. When the nodes

2,5 & 6 receive the RREQ, they check their route to the D. In case

a route to the D is not avail they fwd it to their neighbors. Here

nodes 3, 4 and 10 are neighbors of nodes 2,5 and 6. This is with

the assumption that the nodes 3 & 10 have routes to the D node 15

that is thro paths 10-14-15 & 3-7-9-13-15 resp.

If the Dseqno at node 10 is 4 and is 1 at intermediate node 3 then

only node 10 is allowed to reply along the cached route to S. when

a path breaks for ex bet nodes 4 and 5, both nodes initiates RERR

msg to inform their end nodes abt the link breaks

32

AODV

the end nodes deletes the corresponding entries from their

tables. The source node reinitiates the path finding process

with the new BcastID and the previous Dseqno

Advantages

routes are estab. On demand and Dseqno are used to identify

the latest path

route set up delay is less

disadvantages

Multiple RREP in response to a RREQ packet can lead to a

heavy ctrl overhead

periodic beaconing leads to unnecessary BW consumption

34

Temporally ordered Routing Algor.

Source initiated on demand routing algo which provides

loop free routes to the D

each node maintains its one-hop local topology and also

has the ability to delete partitions

distance metric used in TORA is length of path or height of

node N from the D

3 functions: establishment, maintaining and erasing routes

route estab is performed only when a node requires a path

to a D but doesn’t have any directed link

this process estab D oriented Directed Acyclic Graph

(DAG) using query/update mech.

35

Temporally ordered Routing Algor.

When a node has a data packet to send to D node 7 it sends

query packet. This query packet is fwd by intermediate nodes 2,

3, 4, 5 & 6and reaches D node 7 or any other node which has

route to D node.

When the query packet reaches D, it sends reply containing its

distance from D.

Each node that receives the update packet sets its distance to a

higher value than the distance of the sender of the update

packet. By doing this, a set of directed links from the node

which originated the query to the D node 7 is created.

When an intermediate node(5) discovers that the route to the D

is invalid, it changes its distance value to a higher value than its

neighbor and originates an update packet.

37

TORA

The neighbour node 4 that recives the update packet

reverses the link b/n 1 and 4 and forwards the update

packet. This is done to update the DAG

corresponding to D node 7.

Advantages

by limiting the ctrl packets for route reconfigurations

to a small region, TORA incurs less ctrl overhead

Disadvantages

the local reconfiguration of paths results in non-

optimal routes

38

Location Aided Routing

Uses the location info for improving the efficiency of routing by

reducing the ctrl overhead

availability of GPS for obtaining the position info necessary for

routing

LAR designates two regions for selective fwd of ctrl packets

namely

1. Expected Zone

region in which the destination node is expected to be present

given info regarding its location in the past and its mobility info

2. Request Zone

geographic region within which the path finding ctrl packets are

permitted to be propagated

39

LAR 1 & 2 LAR uses flooding but here flooding is restricted to a small

geographical region

the node forward or discard ctrl packets based on two algo namely

LAR1 and LAR2

The source node explicitly specifies the req zone in RREQ packet

as per LAR1 the RZ is a small rectangle that includes src and dest

nodes sides of which are parallel to x and y axis when S is outside

the EZ

when S is inside the EZ the RZ is reduced to EZ

the src node (node 1) originates the RREQ which is broadcast to

its neigh(2,5,6)

nodes 2 and 6 forwards the RREQ & node 5 discards the RREQ

because it is outside the RZ

40

LAR 1 & 2 finally the RREQ reaches the dest (node 8) which orginates route

reply that contains current location and current time of the node

the src node uses these info for route establishment

the src node (node 1) includes the distance b/w itself and the dest

node (node 11) along with (x,y) coordinates of D in the RREQ

packet

when the intermediate node receives this RREQ packet it

computes the distance b/w itself to D node.

If this distance is less than the distance from S to D+* where * is a

parameter of the algo decided based on the err in location

estimation and mobility then the RREQ packet is fwd. Otherwise

RREQ is discarded

42

LAR 2

Node 5 sends the RREQ this is received by nodes 1,2,4,7 and 6

only nodes 4 & 7 forwards the RREQ

other nodes 1,6,2 discards the RREQ because the distance b/w these

nodes and the D node is greater than the distance b/w S node and D

node

once the RREQ reaches the D(node 11) it generates and send RREP

which contains the path thro which future data packets are to be

propagated

Advantages

LAR reduces the ctrl overhead by limiting the search area for

finding a path

Disadvantages

protocol cannot be used in place where GPS access is not possible

44

Signal-Stability based Adaptive RP

SSA is an on demand routing that uses the signal

stability as a prime factor for finding stable routes

it is beacon-based in which the signal strength of

beacon is measured for determining link stability

protocol consist of 2 parts

Forwarding Protocol(FP)-performs actual routing to

forward a pack on its way to the D

Dynamic RP(DRP)- uses an extended radio interface

that measures the signal strength from beacons - it

maintains a rt by interacting with DRP processes on

other hosts

45

Signal-Stability based Adaptive RP Every node maintains a table that contains the beacon

count and the signal strength of each of its neighbors

if the node has received strong beacons for the past few

beacon the node classifies the link as strong/stable link

the link is otherwise classified as weak/unstable link

a src node which doesn’t have a route to the D floods the

n/w with RREQ pack

The nodes that employ SSA protocol process a RREQ

only if it is received only if it is received over a strong

link

a RREQ received thro weak link is dropped without

processing

46

SSA the dest. Selects the first RREQ and initiates RREP packet to notify the

selected route to the S

when a link breaks the end nodes of the broken link notify the corresponding

end nodes of the path. A src node rebroadcasts the RREQ to find another

stable route. If no strong path is available when a link gets broken then the

new route is estab. By considering weak links also.

Advantages

It finds more stable routes when compared to the shortest path route selection

protocols such as DSR and AODV

Disadvantages

it puts a strong RREQ fwd condition which results in RREQ failures

a failed RREQ reinitiates a path find process without considering stability

criteria - BW is consumed

strong links criterion increases the path length as shortest paths may be

ignored for more stable paths

48

Zone Routing Protocol

Hybird rp which effectively combines the adv of both proactive

and reactive

proactive - Intra zone RP(IARP)- for nodes within a particular

zone

Reactive - Inter zone RP(IERP) - for nodes beyond this zone

the routing zone of a given node is a subset of the n/w within

which all nodes are reachable within less than or equal to zone

radius hops

within routing zone each node maintains the info abt the routes

to all nodes by exchanging periodic route update packets

IERP is responsible for finding paths to nodes which are not

within the routing zone

49

Zone Routing Protocol

when a node S(8) has packet to be sent to node D(16) it checks

whether D is within its zone. If the dest. Belongs its own zone then it

delivers the pack directly.

Otherwise node S bordercast(uses unicast routing to deliver pack

directly to the border nodes) the RREQ to its peripheral

nodes(2,3,5,19,14,15).

If any peripheral finds a path to node D then it sends RREP

otherwise it rebordercast the RREQ. This process continues until D is

located.

Nodes 10 and 14 find the info abt 16 therefore they send RREP pack

back to node 8.

When an intermediate node in an active path detects a broken link in

the path it performs a local path reconfig. In which broken link is

bypassed by means of a shorter alternate path

51

ZRP

Advantages

reduces ctrl overhead compared to the RREQ

flooding mechanism employed in on-demand

approaches and the periodic flooding of routing info

in table driven approaches

Disadvantages

the decisions on the zone radius has a significant

impact on the performance of the protocol

52

Power Aware Routing Protocol

Power consumption by the nodes is a serious factor to be taken

into consideration by RP for AWN

the routes are also equally power constrained just as the nodes are

power aware routing metrics

singh et al. Proposed a set of routing metrics that supports

conservation of bat power

1. Minimal energy consumption per packet - min the power

consumed by a packet in traversing from S to D. The energy

consumed is the sum of energies required at every intermediate

hop in that path. The energy consumed at intermediate path is a

fun. Of distance b/w the nodes that form the link and load on the

link

53

Power Aware Routing Protocol 1. q

2. Maximum n/w connectivity - balancing the routing load among

the out-set

3. Minimum variance in node power levels - to distribute the load

among all nodes in the n/w so that the power consumption pattern

remains uniform across them

4. Minimum cost per packet - in order to max the life of every node

in the n/w, this routing metric is made as a fun. Of the state of the

nodes bat. A nodes cost decreases with an increase in its bat

charge and vice versa

5. Minimize maximum node cost - min the max cost per node for a

pack after routing a number of packets or after a specific period.

This delays the failure of a node occurring due to higher

discharge because of pack forwarding

54

Transport layer

in Ad Hoc Networks

55

Transport layer, Security protocols

The objectives of transport layer include setting up of end to

end connection, end to end delivery of data packets, flow

control and congestion control

Example

UDP-Simple, Unreliable, Connectionless

TCP- Reliable, Byte stream based, Connection oriental.

These traditional wired transport layer protocols are not

suitable for adhoc wireless n/w due to inherent problems

associated with the latter.

The adhoc networks are highly vulnerable to security compared

to wired networks. Therefore, security protocols used in other

n/w cannot be directly applied to adhoc wireless networks.

56

Issues of designing a TL protocol

Induced traffic

The traffic at any given lines due to the traffic in neighboring lines is referred as

induced traffic. This induced traffic affects the throughput achieved by the transport

layer protocol.

Induced throughput unfairness

This refers to throughput unfairness at the transport layer due to the throughput/delay

unfairness existing at the lower layers such as the network and MAC layer.

Separation of congestion control, flow control and reliability

The reliability and flow control are end to end activities whereas the congestion at

time can be local activity. The performance of the transport layer protocol can be

improved if these are handled separately.

Power and BW constraints

The performance of a transport layer protocol is significantly affected by there

constraints.

57

Issues of designing a TL protocol

Misinterpretation of congestion control

Packet loss occurs in adhoc n/w due to high error rates of channel, hidden

terminal problem etc. This may lead to misinterpretation of congestion.

Completely decoupled transport layer

Wired n/w transport layer protocols are completely decoupled from the lower

layers. In adhoc n/w, the cross layer interaction b/n transport layers and lower

layers is important for transport layer to adopt to the changing environment.

Dynamic topology

Some of the deployment scenarios of adhoc wireless networks experience

rapidly changing n/w topology due to mobility of nodes. Hence, the

performance of transport layer protocol is affected by the rapid changes in n/w

topology.

58

Design goals of a TLP

Should maximize the throughput/connection

Should provide throughput fairness across contending flows

Should have mechanisms for flow control and congestion

control

Should be able to provides both reliable and unreliable

connections as per the req.

Should incur min connection set up and connection

maintenance overhead

Should minimize the resource requirements for setting up

and maintaining the connection

59

Design goals of a TLP

Should be able to adopt to the dynamics of the n/w

BW must be used efficiently

Should be aware of resource constraints such as battery

power and buffer sizes and make efficient use of them

Should make use of the information from the lower layers

for improving networks throughput

Should have well defined cross layer interaction framework

for effective scalable and protocol independent interaction

with lower layers

Should maintain end to end

60

Why does TCP not program well in AN

1.Miss interpretation of packet loss

2.Frequent path breaks

3.Effect of path length

4.Miss interpretation of congestion window

5.Asymmetric link behavior

6.Unidirectional path

7.Multipath routing

8.Network partitioning and reemerging

9.The use of sliding window based transmission.

62

Feed back based TCP (TCP-F)

TCP-F is a feedback-based

approach. It requires the support of

a reliable link layer and routing

protocol that can provide FB to the

TCP sender about the path break.

The routing protocol is expected to

repair the broken link within a

reasonable period.

A TCP session is setup b/n node A

and D over the path A-B-C-D

When the intermediate link b/n

node C and node D fails, node C

originates the RFN packet and

forwards it on the reverse path to

the source node.

63


Senders TCP stale is changed to snooze stale upon the receipt of

RFN packet.

In snooze stale a sender stops sending anymore packet to the

destination cancels all the timers, freezes its congestion window,

freezes the retransmission timer and sets up a route failure timer.

When route failure expires the TCP sender changes stale from

snooze stale to connected state.

If the link CD rejoins or if any of the intermediate nodes obtains

a path to destination node a RRN packet is sent to node A and the

TCP state is updated bark to connected state.

64


Advantages

TCP-F provides a simple FB based solution to minimize the problem arising

out of frequent path breaks in adhoc wireless networks.

At the same time, it also permits the TCP congestion control mechanism to

respond to congestion in the n/w.

Disadvantages

If the route to the sender is not available at the FP then additional control

packets may need to be generated for routing the RFN packet.

TCP-F has an additional state compared to the traditional TCP state m/c, and

hence its implementation requires modifications to the existing TCP libraries.

Congestion window used after a new route is obtained may not reflect the

achievable transmission rate to the n/w and the TCP-F receiver

65

TCP with Explicit Link Failure Notification -ELFN

The ELFN is originated by the node detecting a path break upon

detection of a link failure to the TCP sender. This can be

implemented in two ways:

By sending Internet Control message protocol (ICMP) destination

unreasonable (DUR) message to the sender.

By piggy barking this information on the Route Error message

that is sent to the sender.

Once the TCP sender receives ELFN packet it disables its

retransmission timers and enters a standby state. In this state it

periodically originates probe packets to see if a new route is

reestablished.

Upon reception of ACK by the TCP receiver for probe packets, it

leaves the standby state, restores the retransmission timers, and

continues to function as normal.

66

TCP with Explicit Link Failure Notification -ELFN

Advantages

TCP-ELFN improves the TCP notification performance by

decoupling the path break information from congestion

information by the use of ELFN.

It is less dependent on routing protocol and requires only link

failure notification about the pat break.

Disadvantages

When the n/w is temporarily partitioned the path failure may last

longer and this can lead to the origination of periodic probe

packets consuming BW and power.

The congestion window used after a new route is obtained may

not reflect the achievable transmissions rate acceptable to the n/w

and the TCP receiver.

67

TCP-Bus TCP with buffering capacity and sequence information (TCP-Bus) is

similar to TCP-F and TCP-ELFN in its use of feedback information from

an intermediate node on detection of a path break. TCP-Bus was

proposed with Associativity bared routing (ABR) scheme. TCP-Bus

works as follows.

Upon detection of a path break an upstream intermediate node (called

pivot node PN) originates explicit route disconnection notification

(ERDN) message.

This ERDN is propagated to the TCP-Bus sender

Upon reception of ERDN, the TCP-Bus sender stops transmissions and

freezes all times and windows.

The packets in transit at the intermediate nodes from TCP-Bus sender to

PN are buffered until a new partial path from the PN to the TCP-Bus

receiver is obtained by PN

68

TCP-Bus Upon detection of a path break, the down stream node originates

the route notification (RN) packet to the TCP bus receiver

PN attempts to final an alternate route to the TCP-Bus receiver

and availability of such partial route is to destination is intimated

to the TCP-Bus sender through an explicit route successful

notification (ERSN) packet.

The Local Query (LQ) packet carries the sequence number of the

segment at the head of the queue buffered at the PN and REPLY

carries the sequence number of the last successful segment the

TC-Bus receiver received. This enable the TCP-Bus receiver to

understand the packets lost in transition and those buffered at the

intermediate nodes.

The Lost packets are retransmitted by the TCP-Bus sender.

70

TCP-Bus

Advantages

Performance improvement and avoidance of fast

retransmission due to the use of buffering, seq

numbering and selective acknowledgement.

It takes advantage of ABR

Disadvantages

Increased dependency on the routing protocol and

buffering at intermediate nodes

The failure of intermediate nodes that buffer the packets

may lead to loss of packets and performance

degradation

71

Ad Hoc TCP (ATCP)

ATCP also uses the feedback mechanism to make the sender aware

of the status of the network path. Based on the feedback information

retrieved from the intermediate nodes, the TCP sender changes its

state to the persist state, congestion control state, or the retransmit

state.

72

The major functions of the

ATCP layer is to monitor the

packets sent and received by the

TCP sender, the state of the

TCP sender, and the state of the

network.

ATCP The four states in the ATCP are (I) NORMAL (II) CONGESTED (III) LOSS

(IV) DISCONN

When a TCP connection is established, the ATCP sender is in NORMAL

State. In this state, ATCP does not interfere with the operation of TCP

When packets are lot or arrive out-of-order at the destination, it generates

duplicate ACKs. In traditional TCP, upon reception of duplicate ACKs, the

TCP sender invokes the congestion control. But the ATCP sender counts the

number of duplicate ACKs received, if it reaches three, of it puts TCP in

persists state and ATCP in loss state

When a new ACK comes from TCP receiver. It is forwarded to TCP and the

TCP sender is removed from the persists state and then the ATCP sender

change to the NORMAL state.

When ATCP sender is in loss state, the receipt of an ECN message charges it

to the CONGESTED State. Along with this transition, ATCP sender removes

the TCP from the persists state.

When the n/w gets congested, the ECN flag is set in the data and the ACK

packets.

73

ATCP

When ACTP sender receives this ECN message in the normal state, it

changes to the CONGESTED State, and permits TCP to invoke

congestion control mechanism.

When a route failure or n/w partition occurs in the n/w, the n/w layer

details these and informs to the ATCP sender through DUR message.

Upon reception of DUR message, ATCP puts the TCP sender in

persists state and enters into DISCONN state

It remains in the DISCONN state until it is connected and receives

data or duplicate ACKs

On the occurrence of any of three events, ATCP changes to the

NORMAL State.

The receipt of DUR message in the LOSS state or CONGESTED

state causes a transition to the DISCONN state

74

ATCP

Advantages

It maintains the end to end

semantics of TCP

It is compatible with traditional

TCP

Disadvantages

The dependency on the networks

layers protocol to detect the route

changes and partitions, which not

all-routing protocols may

implement.

The addition of thin TCP layer to

the protocol stack that requires

changes in the interface functions

currently used.

75

Split TCP

In networks, the short connections generally obtain much higher

throughput than long connections. This can also lead to unfairness

among TCP session, where one session may obtain a much higher

throughput than the other sessions.

Split TCP provides the solution to the throughput unfairness problem by

splitting the transport layers objectives into congestion control and end

to end reliability. The congestion control is mostly a local solution. At

the same time, reliability is an end to end requirement and needs end to

end acknowledgements.

The operation of the split TCP is shown in fig where a three stage split

connection exists b/n node 1 and node 15.

A proxy node receive the TC packet, reads it, store it in its local buffer

and sends LACK to the source ( or the previous proxy)

76

Split TCP

The responsibility of further delivery of packets is assigned to

proxy node. A proxy node clears a buffered packet once it

receives LACK from the immediate success or proxy for that

packet

the split TCP maintains the end to end ACK mechanism in

addition to zone wise LACK.

The source node clears the buffered packets only after receiving

the end to end acknowledgements.

77

Split TCP

Advantages

improved throughput

improved throughput fairness

Lessened impact of mobility. Since in split TCP the path length

can be shorter than the end to end path length, the effect of

mobility on throughput is lessened.

Disadvantages

It require modifications to TCP protocol

The end to end connection handling of traditional TCP is

violated

The failure of proxy nodes can lead to the throughput

degradation.

79

Application Controlled TP (ACTP)

ACTP assigns the responsibility of ensuring reliability to the application

layer. It is more like UDP with FB of delivery and state maintenance. ACTP

stands b/n UDP and TCP It is not an extension of TCP.

As shown in fig the application layer uses API functions to interact with the

ATCP layer. Each API function sends a packet (Send To ( ) ) to the ACTP

layer which contains information such as maximum delay the packet can

tolerate the message number of the packet and the priority of the packet. The

ATCP also maintains the delivery status through another API function. Is

ACKed <message number> and is available for application layer. A zero in

the delay field refers to the highest priority packet, which requires immediate

txn with min delay.

80

Application Controlled TP (ACTP)

Advantages

It is scalable for large n/w

throughput is not affected by path break as much as

in TCP.

It provides freedom of choosing the required

reliability level to the application layer.

Disadvantages

It is not compatible with TCP.

81

Ad hoc transport protocol (ATP)

ATP is specifically designed for AWN and is not a variant of TCP. The major

aspects by which ATP defers from TCP are

Coordination among multiple layers

Rate based transmission

Decoupling congestion ctrl and reliability

Assisted congestion ctrl

ATP uses information from lower layers for

Estimation of the initial transmission rate

Detection, avoidance and control of the congestion

Detection of Path breaks

The intermediate nodes attach the congestion information to every ATP packet

and the ATP receiver collects it before including it in the next ACK packet. The

congestion information is expressed in terms of the weighted average queuing

delay (DQ) and contention delay (DC) experienced by the packet.

82

ATP

During a connection setup process or when ATP recovers from a path breaks,

the txn rate to be used is determined by a process called quick start. During

quick start process, the ATP sender propagates a probe packet to which the

intermediate nodes attach the transmission rate, which is received by ATP send

receiver, and an ACK is sent back to ATP sender. The ATP sender starts using

the newly obtained transmission rate by setting the data transmission timers.

After congestion occurs, ATP controls it using three phases, namely, increase,

decrease, and maintain

If R>S(1+r) then the current txn rate is increased by a factor k. where R->

new txn rate S->current txn rate r->threshold k->difference b/n new txn

rate and current txn rate

If the new txn rate is higher than the current transmission rate but less than

the threshold then the current txn rate is maintained.

If an ATP sender has not received any ACK packets for two consecutive

feedback periods it undergoes a multiplicate decrease of the txn rate.

83

ATP After a third period without any ack the connection is

assumed to be lost and the ATP sender goes to the

connection initiation phase during which it periodically

generates probe packets

Advantages

Improved performance

Decoupling of the congestion control and reliability

mechanism

Avoidance of congestion window fluctuations

Disadvantages

Lack of interoperability with TCP

84

Questions ?

cs6003 ad hoc and sensor networks

Engineering