Implementation of P2P Computing in Design of MANET Routing Protocol

Zupeng Li
1
, Xiaochuan Yin
1
, Peiyang Yao
1
, and Jinnan Huang
2
1
Telecommunication Engineering Institute, Airforce Engineering University
No. 1 Feng Gao Road, 710077, Xi'an, Shanxi, P.R. China
2
Lenovo Corporate Research & Development, Lenovo China
No. 6 Shang Di West Road, Haidian District, Beijing, P.R. China
Zupengli@hotmail.com
Abstract
The study of peer-to-peer network and mobile ad hoc
network (MANET) are currently two hotspots in distributed
computing and mobile communication researching domain.
By building up a P2P overlay network on top of MANET's
physical infrastructure, we effectively integrated P2P
networks advantage on sustaining highly dynamic network
into the design of MANET routing protocol. By deploying
passive MANET routing algorithms with innovative P2P
routing mechanisms, we propose a new kind of MANET
routing protocol named Peer Computing based Dynamic
Source Routing (PDSR). A detailed description of the P2P
decentralized naming, route discovering, route querying
and updating algorithm used in PDSR is presented in this
paper. The simulation results indicate that PDSR has an
improved routing performance in comparison with the
popularly used DSR protocol.
Keywords: Detour, Hashing, MANET, Peer-to-Peer,
Structured P2P systems
1. Introduction
Peer-to-peer (P2P) network is currently a hotspot in the
international network researching domain [1]. Recently,
different kinds of P2P network models such as Napster,
Freenet and Gnutella have already been popularly used.
However, each of them has some open problems [2] [3]. To
solve these problems, structured P2P overlay networks
(SON) such as CAN, Chord [4], Pastry and Tapestry are
proposed to provide a kind of mechanism to improve
network scalability [5]. All of these SON algorithms have a
common ground that they all assign keys to data and nodes
through a hashing function, and at the same time they use
these keys to store and maintain data which are shared in the
network [6].
On the other side, nowadays mobile ad hoc network
(MANET) technology also draws great attention of
worldwide researchers and scientists. Since the first
appearance of wireless ad hoc networks in the DARPA
packet radio networks in the 1970s, it becomes an
interesting research object in the computer industry. During
the last couple of years tremendous improvements are made
in the research of ad hoc networks. Due to its possibility to
create and organize a network without any central
management, MANET is characterized as the art of
networking without a network [7].
Currently, the research of P2P network and MANET are
always taken as two separated areas. But the results of our
study indicate that there are lots of joints between these two
technologies.
In Section 2, we analyze the synergy of P2P network and
MANET from different aspects. By doing this, Section 3
provides a new routing protocol for MANET named PDSR,
which efficiently integrates P2P computing with MANET
routing algorithms. Optimizations are introduced in section
IV for PDSR to overcome the popularly detouring problem
in structured P2P systems. The Simulation is made in
section V by comparing the overall routing performance of
PDSR with that of DSR. Finally the conclusion is achieved
in Section 6.
2. Similarities between MANET and P2P
networks
By analyzing the synergy of MANET and P2P networks,
we summarize the similarities between them as follows (see
Figure 1):
A. Dynamic network topology
Nodes in P2P networks may randomly join and leave the
network without any signs, which causes the network
topology changes frequently with time. It is the same with
MANET. And the mobility character of mobile nodes in
MANET can make network topology changes much more
frequently [8].
B. Multi-hop connection
Nodes in both of two networks connect with each other
via multi-hop routing. Hop-to-hop connections in P2P
network are typically via TCP links with physically
unlimited range, while hop-to-hop connections in MANET
Proceedings of the First International Multi-Symposiums on Computer and Computational Sciences (IMSCCS'06)
0-7695-2581-4/06 $20.00 2006 IEEE
are via wireless links, which are limited by the radio
transmission range.
C. Decentralized control
Both P2P network and MANET have a decentralized
structure. And there is no such an administration point in the
network as what is in a Client/Server mode network. Nodes
in the network are all peers and can communicate directly to
each other. Network will not suffer the paralysis caused by
central server, which improves the robustness of network.
D. Nodes multi-identity
Nodes in P2P network and MANET act both as a client
and a server, at well as a router. Each node can route and
transfer messages independently. And they can also provide
resource downloading and communicate with each other
directly [9].
(a). Architecture of P2P Network
(b). Architecture of MANET
Figure 1. Comparison of P2P network and MANET
Besides these, unstructured p2p networks such as
Gnutella share additional similarities with MANET such as
(i) flooding-based routing protocols and (ii) limited
scalability due to bandwidth consuming traffic from
flooding.
3. Design of PDSR Protocol
A. Architecture of PDSR Model
The object of our design is to construct a new type of
MANET routing model based on P2P computing
technology. Figure 2 depicts the architecture of PDSR
model, which can be used for providing file discovery
service, information administration service, file transfer
service and communication security service.
B. Basic design
1) Node naming mechanism
PDSR adopts the similar node naming mechanism used in
Chord and HASN. By adopting a consistent hashing
algorithm [11][12], PDSR assigns each node an m-bit node
identifier (NID) (N
d
2m). NID is chosen by hashing the
node's IP address, which can be used to specify the location
of node in a hash ring. When a node joins the network for
the first time, a NID ranges from 0 to (2
m
-1) will be
automatically assigned to it based on the adopted hashing
algorithm. In contrast to HASN, the relationship between a
key identifier (KID) and a NID is one to one mapping. In
order to simplify the routing algorithm, PDSR adopts the
discontinuous naming mechanism like Chord.
Fig.3 depicts the node naming mechanism used in PDSR,
while Fig.3 (a) depicts the physical structure of MANET
and (b) depicts the corresponding hash ring structure in the
logical namespace.
Figure 2. Architecture of PDSR Model
2) Node state
PDSR routing table (PRT) has a similar structure to that
of the routing table in dynamic source routing protocol. The
difference lies in that the destination node and next hop
node are all depicted by nodes NID in PDSR instead of
their IP addresses. The following concepts are related to a
mobile node's PRT in an m-bit namespace.
Predecessor: the first node in the counterclockwise
direction of hash ring from current node;
Successor: the first node in the clockwise direction of
hash ring from current node;
PRTEntry[K].start: the starting location of the ring range
File
Discovery
Service
Information
Administration
Service
File
Transfer
Service
Communication
Security Service
P2P Computing Based MANET Routing Algorithm
P2P Overlay Network (Node Naming and Indexing)
MANET Physical Network
Proceedings of the First International Multi-Symposiums on Computer and Computational Sciences (IMSCCS'06)
0-7695-2581-4/06 $20.00 2006 IEEE
that the K
th
PRT entry covers.
1
[ ]. ( 2 )mod2 1 (1)
k m
HPRTEntry K start n K m

d d
Next Hop Ring Interval (NRI): the ring range that
the Kth PRT entry covers.
[ ] [ [ ]. , [ 1]. )
1 K (2)
NRI k PRTEntry K start PRTEntry K start
M

d d
Next Hop Identifier (NHID): the NID of the first active
node G in the clockwise direction of NRI.
[ ]. ( ) 1 (3) PRTEntry K start NID G K M o d d
Next Hop Source Route (NHSR): the source route to the
corresponding next hop node G in PRTEntry[K].
Suppose current node is S
1
, NHSR satisfies the
following formula.
) 4 ( ,..., , ,
3 2 1
! G S S S NHSR
A PRT is composed of PRTEntry[K].start, NHID and
NHSR. Fig.4 depicts the structure of PRT.
PRTEntry[K].start NHID NHSR
< PRTEntry X> < NHID Y> <S
1
S
2
S
x
>
Figure4. The structure of PRT
C. PDSR route discovering algorithm
The function of PDSR route discovering method is to
discover the source route to destination node reactively.
Generally, PDSR route discovering method is triggered by
mobile nodes route updating method under the following
status:
i. When a node initiates a route query procedure or has
received a route query message (RREQ), it will first query
its PRT. If the corresponding route entry is null or outdated,
it will trigger route discovering method.
ii. When a middle node in the source route returns a route
error message (RERR) to the node which initiates the data
packet, the initiating node will start the related route
updating method, which will then trigger the route
discovering method.
PDSR route discovering algorithm can be depicted as
follows:
(a). The structure of physical MANET
Nl D=0001
B
A
E
F
Nl D=0100
Nl D=0000
Nl D=1100
0 2
4
Nl D=0010
D
Nl D=1000
C
(b). The structure of hash ring in logical namespace
Figure 3. PDSR Naming Mechanism
1). When a node M initiates or receives a RREQ message,
it will firstly compare the source id and request id of the
message with those RREQ it has cached. If they are
matched, it means the RREQ message has been received
already, M will then discard it. Otherwise, go to step 2;
2). M will check the Route_record of RREQ message to
verify whether it contains NID(M). If it does, M will discard
the message. Otherwise, go to step 3;
3). M will check whether destination node belongs to the
PRT entries. If it does, go to step 5; otherwise, go to step 6;
4). M will check whether destination node is M itself. If it
is, go to step 5; otherwise go to step 6;
5). M will check whether the communication channel is
bi-directional. If it is, M will return route reply message
(RREP) to the query initiating node S directly. Otherwise,
M will trigger route discovering method to discover source
route to S and then return RREP to S;
6). M will add itself into the Route_Record, and broadcast
the RREQ message to its neighbors.
D. PRT querying algorithm
PRT querying algorithm can be depicted as follows:
1). When a node M initiates or receives a RREQ
message, it will firstly compare the destination node id of
RREQ with NID(M). If they are matched, M will return
RREP to querying initiating node S. Otherwise, go to step 2;
2). M will search into the PRT and compare destination
node id (D) with PRT entries to find the NRI which D
belongs to. If the PRT entry is found, M will compare D
with the corresponding NHID. If D > PRTEntry[i]. NHID
then transfer the message to the corresponding next hop
node. Otherwise go to 3.
A
C
F
D
E
B
Proceedings of the First International Multi-Symposiums on Computer and Computational Sciences (IMSCCS'06)
0-7695-2581-4/06 $20.00 2006 IEEE
Table 1. PDSR route discovering algorithm
Ddestination node id
When mobile M initiates a Route_Request or receives
Route_Request_Packet {
//To avoid processing duplicate packets
if (<source NID, req_id>

(Taken RREQ[ ])) then{

discard the packet;}
//To make the algorithm loop free
else if (NID(current_node)

Route_Record[ ]) then{
discard the packet;}
else {Pick up D from RREP packet;
// Optimized method: return RREP directly from PRT
if (D

NHID [ ]) then{ return RREP_Hit;}

else if (D == NID(current_node)) then{
if (the link is bi-directional) { return RREP_Hit;}
else { Create Route_Request packets to find Sender;}}
//M is between Sender and Destination
else {
append NID(M) in Route_Record[ ] and broadcast RREQ
Packets;}}
}
3). M will trigger route discovering method for
destination node D and wait for response. If the source route
to D is returned, M will update the NHID and NHSR of the
PRT entry with D and source route to D, respectively. At the
same time, M will return a RREP message to S. Otherwise,
M will return a query failure message to S.
4). During the process, each time a node M queries its
route table, it will pick up the PRT entry whose NRI is
closest to D. And by comparing D with the corresponding
NHID, M will determine whether it should trigger the route
discovering process for D. If it is unnecessary to trigger a
route discovering process, M will transfer the RREQ
message to next hop node in the PRT entry. This process
goes on until a RREP is returned.
To sum up, PRT querying algorithm performs the
following two functions:
i. Route the RREQ message according to the PRT
cached by each node.
ii. Update the outdated PRT entries cached by each node
reactively.
E. PDSR route maintenance algorithm
1) Mobile Node Join and Depart
Due to the highly dynamic character, mobile nodes may
join and leave the network at any time. To save the route
maintenance cost, when a node joins the network, PDSR
will not initialize its PRT. On the contrary, the node will
construct its PRT on demand.
The node join procedure can be depicted as follows:
i. Node M will firstly choose a member node B as its
bootstrap node. And B must satisfy the following criterion:
A) B is inside M's transmitting scope, or we can say B is a
neighbor of M; B) Among all M's neighbors, B has a NID
which is closest to M's.
ii. M will find its successor S on hash ring by the help of
B.
iii. M will get information of S's predecessor P. Then M
will send a join message to both P and S. when they receive
the message, P will update its successor with M and S will
update its predecessor with M. Till now, the integrity of
hash ring is guaranteed.
iv. M will then construct and update its PRT entries
reactively. And the triggering event is the corresponding
PRT entry querying procedure.
v. PDSR classifies node departure event into two types:
vi. Normal departure: nodes leave the system
deliberately.
vii. Abnormal departure: nodes leave the system without
any signs.
To reduce route maintenance overhead, PDSR takes no
action for a mobile node's abnormal departure. As for a
nodes normal departure, PDSR will take the following
actions:
The quitting node will firstly send a QREQ to its
predecessor P and successor S. When S and P receive the
message, S will update its predecessor as P and P will
update its successor as S. Till now, the integrity of hash ring
is guaranteed.
Table 2. PRT querying algorithm
PRTEntry[i]the ith entry in PDSR routing table (PRT) 1
d
i
d
M
Ddestination node id
When mobile M initiates a Route_Request or receives
Route_Request_Packet
{
Pick up D from RREP packet;
if (D

NID(current_node)) then {
return RREP_Hit; }
else { //search in PRT
for k=1 upto M {
if ((D

[ PRTEntry[i].start, PRTEntry[i+1].start ) & (D > NHID [i])) {

forward to NHID [i];}
else{ Invoke Route Discovery for D;
Wait for Reply;
If (Source route to D is Found)
{
Update PRTEntry[i].NHID with NID(D);
Update PRTEntry[i]. NHSR with Source route to D;
return RREP_Hit;}
else{ return RREP_Failed;} } }
} }
2) PDSR route updating algorithm
As a MANET routing protocol, PDSR must handle the
problem of nodes mobility. In MANET, a nodes position
is changing with time, which leads to the dynamic network
topology. When a destination node is out of reach, this may
be caused by some corresponding nodes normal or
abnormal departure, or some nodes have move out of
transmitting scope of others. Once this happens, PDSR will
trigger routing updating algorithm reactively.
Generally, there are two kinds of triggering event:
Proceedings of the First International Multi-Symposiums on Computer and Computational Sciences (IMSCCS'06)
0-7695-2581-4/06 $20.00 2006 IEEE
As for a mobile nodes normal departure, it will
broadcast a QREP to its neighbors. The node which has
received QREP will act as the triggering node of route
updating process.
As for a data packets transferring failure, which will
be caused by nodes abnormal departure or movement, the
node who initiates the data packet will act as the triggering
node of route updating process.
PDSR route updating algorithm can be depicted as
follows:
i. As for condition II, the node N who initiates the data
packet will broadcast a route overdue message into the
network. And those related nodes will delete the outdated
PRT entries accordingly. Then go to step 2. As for condition
I, go to step 4.
ii. N will trigger a route discovering process for the
destination node. If the destination node is found and a
RREP is returned, N will update the outdated source route
in PRT with the currently discovered one.
iii.If the route discovering process ends with a failure
result, this denotes that the destination node may be down or
has already left the network. In such a condition, N will
terminate the route discovering process and mark the
destination node as invalid. Then go to step 4.
iv. The triggering node will broadcast the destination
node invalid message into the network, the predecessor and
successor of the invalid node will update their successor and
predecessor after they receive the message, respectively.
Till now, the integrity of hash ring is guaranteed. Other
related nodes will then delete the outdated entries from their
local PRT after they receive the message.
4. PDSR Optimizations
F. Overhearing
One of the most important optimizations in DSR is the
introduction of overhearing [10]. In DSR, each node can
obtain route information by overhearing data packet or
RREP packet passing by it. As for MANET depicted in
figure 1, if D receives a RREP message which B sends to A,
D can obtain its source route to A by overhearing the source
route [B,E,D,F,C,A] in RREP. In the mean time, mobile
nodes can also obtain the route information by overhearing
the data packet passing by it.
In order to decrease the network maintenance cost, PDSR
has also introduced overhearing technology. Each node will
actively overhear the packets passing by it. But in contrast
to DSR, PDSR nodes will only overhear the source route
related to the next hop node in its local PRT. By doing this,
PDSR inherits all the advantages of DSR and at the same
time limits the related nodes to be overheard to O(logN).
Consequently, the overload of nodes in PDSR is reduced.
G. Solutions to detouring problem
Detouring problem is a serious problem for DHT
(distributed hash table) based P2P routing algorithms. Since
those routing algorithms have not taken networks physical
topology into consideration, messages between two
physically closed nodes may have to run across a long way
in the logical naming space as a result of hashing procedure.
And this decreases the routing efficiency seriously. Now if
we implement structured P2P overlay network technology
into MANET, we will face the same problem. By analyzing
MANETs characters, this paper proposes two
optimizations for PDSR to solve the detouring problem.
1). Source route detection mechanism
Each mobile node in PDSR keeps a PRT and by analyzing
its PRT entries, a node can relieve the detouring problem to
some extent. The basic idea of source route detection
mechanism is that each time a node initiates or receives a
RREQ message, it will firstly check its PRT entries to find
out whether the destination node exists in the source routes
of PRT entries. If it does, the node will pick up the source
route to destination node directly; otherwise it will perform
the routine PDSR route discovering or PRT querying
algorithm.
As for MANET in figure 1, suppose A initiates a RREQ
for F, by executing routine PRT querying algorithm, A will
send the RREQ to E and E will then transmit the RREQ to F.
But by applying source route detection mechanism, A will
find destination node F in its source route to E by simply
checking the source route <A,C,F,D,E>. Therefore, A can
pick up the source route <A,C,F> directly without
performing the routine PRT querying algorithm or
transmitting the RREQ to other nodes in the network.
2). Phello protocol and PDSR Neighborhood Table
(PNT)
In order to take full advantages of mobile nodes network
proximity character, we introduce a PNT for each node to
keep a record of its neighbor nodes. These neighbor nodes
are those nodes which are inside the transmitting scope of
current node. The structure of PNT is depicted in figure 5.
Sequence number Neighbor Node Id (NNID)
K < NNID Y>
Figure 5. The structure of PNT
Based on PNT, we propose a Phello protocol for PDSR.
The function of Phello protocol is similar to Hello protocol
in AODV. But they are different in implementation. In
Hello, mobile nodes periodically broadcast hello messages
to their neighbors to notify their positions. But in Phello,
mobile nodes broadcast hello messages reactively, which
means only a route request can trigger it. As MANETs
topology changes dramatically with time, a mobile node
Proceedings of the First International Multi-Symposiums on Computer and Computational Sciences (IMSCCS'06)
0-7695-2581-4/06 $20.00 2006 IEEE
neednt keep a record of it neighbor nodes. When a route
request process takes place, the node will then collect its
neighbor nodes information by using Phello protocol. If the
destination node is one of its neighbor nodes, it will stop the
route discovering process and return a RREP directly by
adding destination node into the source route record.
Take figure 3 for example, suppose mobile node M and N
have joined into the network later at a time, and the Physical
and logical structure of current network will change
accordingly, which is depicted in figure 6. And the PRT of
mobile node A will be updated as table 3.
When node A (NID=0001) initializes a route querying
message for node M, it will first query its local PRT and
transmit the querying message to corresponding next hop
node N (NID=0110). Then N will carry on the querying
process. But in fact the physical topology of MANET
indicates that the destination node M is geographically close
to the source node A. Although M is not in the transmitting
scope of node A, it is in the transmitting scope of node F.
Therefore, when route querying message is transferred to
node F according to the source route <A,C,F,D,N>, F can
find out that the destination node F is one of its neighboring
node by simply sending PHello messages. Then F can
construct a query responding message with source route
<A,C,F,M> and return the message directly to node A.
As we can see, by introducing PHello protocol and PNT,
nodes which transmits the querying message can detect and
find out whether the destination node is one of its
neighboring node. By doing this, PDSR can to some extent
avoid the detouring problem which is caused by middle
nodes meaningless message transmission in the logical
namespace.
Table 3. Updated PRT of mobile node A
5. Simulation of PDSR Model
H. Simulation environment
1). Simulation parameters:
Simulation is done with fifty nodes within a
1000m*1000m space. The physical parameters of each
node can be depicted as follows:
Antenna is a full-direction antenna with a height of 1.5
meter, the reception and transmission increments are both
1.0 dB.
The bandwidth of wireless radio is 2 Mb/s, which
adopts a two-path channel model.
The transmission power is 17.6 mw and the radius of
wireless radio broadcasting scope is 110 meters.
2). Mobile model:
Fifty nodes are randomly distributed in the simulation
scope. And each node performs a Brown movement during
the simulation period. Besides these, we adopt the
traditional pause
time (PT) to define the mobile character of our model. At
the beginning of simulation, a node keeps static for a PT.
Then it
will randomly choose a destination node and move toward it
with a certain aped. When it arrives, the node will keep
static for a PT and then choose a node as the next
destination. This process will be repeated till the simulation
ends. We choose for different values of PT: 1, 60, 120, 180
and 360.
3). Service model:
In order to simulate the network services influence on
PDSRs performance, we randomly choose 20 pairs of
nodes to transmit CBR packets to each other. The CBR
packets are 128 bytes packets with a transmission rate of 4
packets per second.
4). MAC layer:
A distributed coordination function (DCF) mechanism is
adopted in MAC layer.
I. Performance results
1). Packet Deliver Ratio (PDR)
Packet deliver ratio is an important criterion in estimating
a routing protocols performance in how well it adapts to the
change of network topology and how efficient it provides a
route discovery. Figure 7 depicts the relationship of PDR in
PDSR, OPDSR (optimized PDSR) and DSR with PT.
(a). The physical structure of MANET after M, N join
sequence
number K
PRTEntry
[K].start
NHID NHSR
1 0010 0010 <A,C,F,D,E,B>
2 0011 0100 <A,C>
3 0101 0110 < A,C,F,D,N>
4 1001 1100 < A,C,F,D,E>
A
C
F
D
E
B
M
N
Proceedings of the First International Multi-Symposiums on Computer and Computational Sciences (IMSCCS'06)
0-7695-2581-4/06 $20.00 2006 IEEE
Nl D=0001
B
A
E
F
Nl D=0100
Nl D=0000
Nl D=1100
0 2
4
Nl D=0010
(b). The logical namespace after M, N join
Figure 6. The physical and logical network structure after M,
N join
As we can see in figure 7, the probability of link failure
between nodes and route failure decreases when the
mobility of nodes decreases. The success rate and stability
of packet transmission in PDSR is much better than those in
DSR. And OPDSR further improves the performance of
packet transmission by keeping the average packet deliver
ratio beyond a rate of 90%. The reason is that PDSR has
improved the node discovery and packet delivery
performance by taking advantages of structures P2P routing
mechanisms. And OPDSR further improves the routing
performance by taking mobile nodes network proximity
character into account and obtaining real time network
status. However, the caching mechanism adopted in DSR is
the origin of its problem in providing out-of-time network
status and its low Packet deliver ratio.
2). Average End-to-end Delay (AED)
Average end-to-end delay is an important criterion in
estimating a routing protocols performance in how
efficient and fast it provides a route discovery. Figure 8
depicts the relationship of AED in PDSR, OPDSR and DSR
with PT.
As we can see in figure 8, due to the broadcasting based
routing discovery mechanism, it takes DSR longer time to
rediscover route when current path is failed. On the
contrary, PDSR improves the route discovery efficiency by
building up a P2P overlay network on MANETs physical
topology and introducing PRT to accelerate the route
discovery process. But due to the detouring problem, the
AED of PDSR is still relatively high. OPDSR further
improves the route discovery efficiency by introducing
optimizing mechanism to avoid detouring problem.
0 50 100 150 200 250 300 350 400
0.72
0.74
0.76
0.78
0.80
0.82
0.84
0.86
0.88
0.90
0.92
0.94
0.96
0.98
1.00
0 50 100 150 200 250 300 350 400
0.72
0.74
0.76
0.78
0.80
0.82
0.84
0.86
0.88
0.90
0.92
0.94
0.96
0.98
1.00
0 50 100 150 200 250 300 350 400
0.72
0.74
0.76
0.78
0.80
0.82
0.84
0.86
0.88
0.90
0.92
0.94
0.96
0.98
1.00
PDSR
OPDSR
P
a
c
k
e
t

d
e
l
i
v
e
r

r
a
t
i
o
Pause time(s)
DSR C
N
Nl D=0110
Figure 7. The comparison of PDR between DSR, PDSR and
OPDSR
3). Node Storage Overhead (NSO)
Node Storage Overhead is an important criterion in
estimating how much extra storage overhead that the
routing protocol puts on each node. Figure 9 depicts the
relationship of NSO in PDSR and DSR with the number of
nodes in the network.
As we can see in figure 9, the cached routes for a mobile
node in DSR are in a direct proportion to the number of
nodes in network. That is to say, as for a MANET with N
nodes, the cached routes for each node is O(N). On the
contrary, by implementing structured P2P routing
mechanism, the relationship of cached routes for each node
in PDSR with the number of nodes is O(logN). By doing
this, PDSR effectively decreases the route storage overhead
for each node, and at the same time it decreases the route
discovery and maintenance overhead related to these cached
routes as well.
4). Routing Overhead (RO)
Routing Overhead is an important criterion in estimating
a routing protocols efficiency. Figure 10 depicts the
relationship of RO in PDSR, OPDSR and DSR with PT. As
we can see in figure 10, the number of route packets in DSR
is huge due to the number of cached routes for each node is
huge as well (the number of cached routes for each node in
DSR has a relationship of O(N) with node number N). The
route maintenance and update processes caused by these
cached routes consume lots of network bandwidth and
decrease the performance of DSR. On the contrary, PDSR
improves the routing performance by introducing structured
P2P overlay network technology and limiting the number of
cached routing within O(logN). And OPDSR further
decreases the routing overhead by introducing optimizing
mechanism to update cached routes without sending out
route discovery packets.
D
Nl D=1000
M
Nl D=0111
Proceedings of the First International Multi-Symposiums on Computer and Computational Sciences (IMSCCS'06)
0-7695-2581-4/06 $20.00 2006 IEEE
0 50 100 150 200 250 300 350 400
50
100
150
200
250
300
350
400
0 50 100 150 200 250 300 350 400
50
100
150
200
250
300
350
400
0 50 100 150 200 250 300 350 400
50
100
150
200
250
300
350
400
A
v
e
r
a
g
e

D
e
l
a
y

(
m
s
)
Pause time (s)
DSR
PDSR
OPDSR
Figure 8. The comparison of AED between DSR, PDSR and
OPDSR
6. Conclusion
Based on the study of synergies between P2P network
and MANET, we propose a new type of MANET routing
protocol named PDSR. By building up a P2P overlay
network on top of MANET's physical infrastructure, PDSR
seamlessly integrates the functions of p2p overlay routing
protocols operating in a logical namespace with those of
MANET routing protocols operating in a physical
namespace. By this means, PDSR has inherited all the
advantages of P2P networking technology in sustaining
highly dynamic networks and maintaining network
scalability. Besides these, PDSR introduces a series of
optimizing mechanisms to solve the detouring problem of
structured P2P overlay network and to further improve
routing efficiency. The simulation results indicate that the
overall routing performance of PDSR is remarkably better
over DSR.
0 10 20 30 40 50
0
100
200
300
400
500
600
700
0 10 20 30 40 50
0
100
200
300
400
500
600
700
0 10 20 30 40 50
0
100
200
300
400
500
600
700
N
o
d
e

S
t
o
r
a
g
e

O
v
e
r
h
e
a
d

(
B
y
t
e
s
)
Node Number
DSR
N
o
d
e

S
t
o
r
a
g
e

O
v
e
r
h
e
a
d

(
B
y
t
e
s
)
Node Number
PDSR
OPDSR
Figure 9. The comparison of NSO between DSR, PDSR and
OPDSR
0 50 100 150 200 250 300 350 400
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
0 50 100 150 200 250 300 350 400
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
0 50 100 150 200 250 300 350 400
1000
2000
3000
4000
5000
R
o
u
t
i
n
g

O
v
e
r
h
e
a
d

(
p
a
c
k
e
t
s
)
Pause time (s)
DSR
PDSR
OPDSR
Figure 10. The comparison of RO between DSR, PDSR, and
OPDSR
References
[1] L. Z. Granville, D. A. Panisson, C. Melchiors, M.
Janilce B. Almeida, "Managing computer networks using
peer-to-peer technologies,", IEEE Communications
Magazine, Vol. 43, No. 10, October 2005.
[2] J. Xu, "On the fundamental tradeoffs between routing
table size and network diameter in peer-to-peer networks,"
IEEE INFOCOM, 2003 - The Conference on Computer
Communications, Vol. 22, No. 1, Mar 2003,
pp. 2177-2187.
[3] S. Sen and J. Wang, "Analyzing peer-to-peer traffic
across large networks," IEEE/ACM Transactions on
Networking, Vol. 12, No. 2, Apr 2004, pp. 219-232.
[4] I. Stoica, R. Morris, D. Liben-Nowell, D. R. Karger, M.
F. Kaashoek, F. Dabek, and H. Balakrishnan, "Chord: A
scalable peer-to-peer lookup protocol for internet
applications," IEEE/ACM Transactions on Networking,
Vol. 11, No. 1, Feb 2003, pp. 17-32.
[5] J. Xu, A. Kumar, and X. Yu, "On the fundamental
tradeoffs between routing table size and network diameter
in peer-to-peer networks," IEEE Journal on Selected Areas
in Communications, Vol. 22, No. 1, Jan 2004, pp. 151-163.
[6] S. Jain, R. Mahajan, and D. Wetherall, "A study of the
performance potential of DHT-based overlays," 4th
USENIX Symposium on Internet Technologies and Systems
(USITS 2003), March 2003.
[7] S. Jiang, Y. Liu, Y. Jiang, and Q. Yin, "Provisioning of
adaptability to variable topologies for routing schemes in
MANETs," IEEE Journal on Selected Areas in
Communications, Vol. 22, No. 7, Sep 2004, pp. 1347-1356.
[8] D. Cavalcanti, D. Agrawal, C. Cordeiro, B. Xie, and A.
Kumar, "Issues in integrating cellular networks, WLANs,
Proceedings of the First International Multi-Symposiums on Computer and Computational Sciences (IMSCCS'06)
0-7695-2581-4/06 $20.00 2006 IEEE
and MANETs: A futuristic heterogeneous wireless
network," IEEE Wireless Communications, Vol. 12, No. 3,
June 2005.
[9] C. H. Foh and B. S. Lee, "A closed form network
connectivity formula for one-dimensional MANETs," ICC
2004 - IEEE International Conference on Communications,
Vol. 27, No. 1, June 2004, pp. 3739-3742.
[10] B.-C. Seet, B.-S. Lee, andC.-T. Lau, "Optimization of
route discovery for dynamic source routing in mobile ad
hoc networks," Electronics Letters, Volume 39, 30, Oct.
2003, pp. 1606 1607.
[11] J. Mischke and B. Stiller, "Efficient protocol
specification and implementation for a highly scalable
peer-to-peer search infrastructure," NOMS 2004 -
IEEE/IFIP Network Operations and Management
Symposium, Vol. 9, No. 1, Apr 2004, pp. 411-424.
[12] A. Duran and C.-C. Shen, "Mobile ad hoc P2P file
sharing," WCNC 2004 - IEEE Wireless Communications
and Networking Conference, Vol. 5, No. 1, March 2004,
pp. 114-119.
Proceedings of the First International Multi-Symposiums on Computer and Computational Sciences (IMSCCS'06)
0-7695-2581-4/06 $20.00 2006 IEEE