Академический Документы
Профессиональный Документы
Культура Документы
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 129
Abstract—This research work is based on the enhancement of Quality of Service (QoS) in Peer-to-Peer (P2P) overlay networks. It reduces
free riding by enhancement of cooperation among heterogeneous peers for the fair share of resources, as well as elevates the efficiency of in-
formation retrieval regarding a particular peer and its resources available and demanded for and from the network respectively. These re-
sources could be peer’s physical attributes such as free storage space, RAM, processor cycles, or it could be any other application based re-
sources, like media files etc. An efficient behavior of the peers in the network is retrieved by introducing cooperating groups (CG) in the
network, where cooperating peers are associated to a particular CG. CG-ID is assigned to those peers who share resources on the basis of
give and take rule. Through CGs it would be far more efficient to search an idol peer waiting for another peer so that both could share each
other’s resources. Heterogeneous Cooperating Group-based Newscast Protocol (HCGNP) has been introduced in the network which gener-
ates cooperating peers in the network. Simulations and results show that the probability of cooperation between the peers, who remain part of
the network, is very high as compared to other cooperation algorithms. Hence the current research introduces an entirely different and an ef-
ficient way of enhancing cooperation between the peers in P2P overlay networks.
—————————— ——————————
1 INTRODUCTION
Peer‐to‐Peer (P2P) has become one of the most widely dis‐ scalability and fault tolerance. P2P overlay networks need
cussed terms in information technology. The term peer‐to‐ an extensive enhancement of features like management of
peer refers to the concept that it is a network of equals peers according to network capacity in the network where
(peers) using appropriate information and communication the peers are joining to share or retrieve resources from
systems where two or more individuals are able to spon‐ other peers [2]. If any peer in the network has some extra
taneously collaborate without needing central coordina‐ resources to share with other peers, then it is quite difficult
tion. In contrast to client/server networks, P2P networks to create an environment of cooperation between the
promise improved scalability, lower cost of ownership, peers. In P2P networks, generally, the request or the avail‐
self‐organized and decentralized coordination of previous‐ ability message for a particular file or service is broadcast‐
ly underused or limited resources, greater fault tolerance, ed to other peers in the network, which incorporates not
and better support for building ad hoc networks. In addi‐ only a load on the network (as every node has to pass on
tion, P2P networks provide opportunities for new user these messages to other nodes in the network) but also
that could be implemented using different usual ap‐ vulnerable to free riding, a behavior in which nodes re‐
proaches [1]. All the peers in the network have portion of ceive messages, but do not want to invest their resources
their resources directly available for the other peers. for forwarding them to neighbors. Hence free riding lacks
P2P overlay networks are distributed systems in na‐ the behavior of cooperation between the peers.
ture, without any hierarchical organization or centralized In P2P overlay networks the nodes connected together
control. Peers form self‐organizing overlay networks, that are enriched with physical resources to be shared among
are overlaid on the Internet Protocol (IP) networks, offer a the nodes, other than files and applications, like RAM,
mix of various features such as robust wide‐area routing Extra storage, Processor cycles, etc. The nodes are con‐
architecture, efficient search of data items, selection of nected with each other, normally on the basis of contents
nearby peers, redundant storage, permanence, hierarchical to be shared. But the physical resources of the nodes some‐
naming, trust and authentication, anonymity, massive times limit the content based file sharing, and other appli‐
cations, as the particular content might be available but its
————————————————
physical parameters restrict the ability of sharing that con‐
S.A, the first author of this research is a student of College of Electrical and
Mechanical Engineering, National University of Sciences & Technology,
tent efficiently. The current research emphasizes on creat‐
Islamabad, Pakistan. ing method for efficient collaboration of nodes for sharing
I.K is a student of College of Electrical and Mechanical Engineering, Na- physical resources like storage space and processor cycles.
tional University of Sciences & Technology, Islamabad, Pakistan, and This collaboration should be such that enhances coopera‐
working as a lecturer at Islamic International University, Islamabad, Paki-
stan. tion among the nodes. The method currently discussed is
M.Y.J is the supervisor of this research and is working as an Associate based on formation of cooperating groups (CGs) which
Dean at College of Electrical and Mechanical Engieneering, National Uni- enhance information retrieval between the nodes. These
versity of Sciences & Technology.
CGs are formed on the basis of the cooperation level asso‐
ciated to them. By cooperation level it means either the
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 5, MAY 2011, ISSN 2151-9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 130
node in the network requires or demands a particular re‐ cache of every node is traversed and the cache list is
source from the other node. The CG formation is an on‐ transmitted to the neighboring nodes through SEND‐
demand procedure depending on whether the node is STATE( ) method. After receiving a message transmitted
cooperating or not. by a node through SENDSTATE( ), the UPDATE( ) proce‐
Hence the current research work is based on utilization dure mixes this received view with the node’s current
of resources in P2P overlay networks. The formation of state view. This merged view is trimmed by the Newscast
P2P overlay architecture is based on sharing resources to obtain the predefined ‘C’ size. Fig. 1 illustrates the pro‐
among different peers. These peers (nodes) join the net‐ cedure adopted by newscast protocol. It must be noted
work in order to gain benefits from the network in the that the most ‘old’ node descriptor is discarded in each
cycle, thus continuously altering the node descriptor hold
form of resources like, different files, applications, band‐
for each node view. This technique updates the overlay
width, extra storage, RAM and processor cycles. When the
defined by a set of all node views.
nodes with variant properties and physical resources join
a network, such a network is known as heterogeneous
network. Sometimes a node demands a resource which it
has shared with other nodes previously [1].
based Newscast Protocol proaches in which this hierarchy is not followed. In case
As P2P overlay computing is a layered approach, there is of P2P overlay networks non hierarchical groups could be
a need to maintain an architecture which should utilize developed which do not have any super peer concept in
network resources with minimum load over the network. the network. As pure P2P overlays are totally decentral‐
Once a fine architecture has been created the throughput ized, the non‐hierarchical concept in the groups or clus‐
of the networks accelerates with the passage of time. Af- ters would help the network to distribute the resource
ter a thorough analysis of previous researches [5] the fol- sharing phenomena to every node participating in the
lowing facts could be extracted for proposing a well de- network. Grouping of the nodes is also helpful to organ‐
fined solution: ize P2P overlay network so that the requests are routed
1. There is a need of an architecture which should more efficiently. Hence the nodes grouped together to
control the peer nodes according the network ca‐ share resources in such a way that every node gains a fair
pacities. share of demanded resources from other peers, could be
2. The node search should be more efficient. an ideal way to eliminate nodal search which is extended
3. The criteria for giving extra benefits to the nodes to every single node in the network. The efficiency of the
should depend upon the type of the network and network to retrieve information regarding the files and
network capacity. services available on the other nodes is one of the most
4. Only those nodes should become the part of the required and demanded property of a particular P2P
network which show cooperating behavior. network. The creation of groups not only minimizes the
5. Nodes that do not cooperate should be discarded time to search the required resources but can also gives a
from the network; this will automatically de‐ free hand to manage the nodes on the application level in
crease the network load. order to generate security protocols like authentication
6. The behavior of a particular node should depend
and to maintain network data and node level information.
upon the capability levels of the node.
7. These capability levels should be according the
node’s physical parameters like presence of extra 3 IMPLEMENTATION OF HETEROGENEOUS
secondary storage, RAM and processing power. COOPERATING GROUP-BASED NEWSCAST
8. The diversified behavior of a node, i.e. it may re‐ PROTOCOL
quires a resource as a demanded physical pa‐
rameter or demanded capability level and anoth‐ In a heterogeneous P2P overlay network existing peers
er resource as available capability level, should posses a heterogeneous nature. The heterogeneous net-
be there to achieve certain level of cooperation work is such that it has peers with different types of op-
with other nodes. erating systems, processors, RAMs, and storage capaci-
9. This diversified nature of the node also reflects ties. Such environment in the network needs a challeng-
the heterogeneous nature of the network which ing architecture to be implemented. Apart from the need
exists with variable values of physical parame‐ of cooperating behavior, problems related to security and
ters. network management are required to be handled proper-
10. The overlay architecture should be such which ly. The objective of P2P overlay network is to share re-
not only incorporates cooperation among the sources in the real time networking environment where
nodes but also handles the network with hetero‐ the nodes (peers) with the heterogeneous properties need
geneous nodal capabilities. to collaborate with each other. The architecture should
11. The nodes which are discarded from the network not only handle the network heterogeneity but also be
could join the network again, with some different
able to incorporate quality of service. This research en-
capability parameters in order to provide a fair
hances the existence of heterogeneity of the peers in P2P
share gain of other’s (nodes) resources.
overlay network and discusses different behavioral as-
12. Every time a node joins a network it becomes
part of cooperating group, a group of nodes pects of the peers accordingly. The role of cooperating
where every node is cooperating with the other nodes in heterogeneous environment is discussed in de-
node. tail where nodes cooperate on the basis of some benefit
13. The approach for development of cooperating offered in the network.
groups should be implemented as to enhance the It has also been discussed that how cooperation can be
efficiency of node search procedure. This could enhanced by development of cooperating groups in a P2P
be true when the nodes search the other nodes overlay architecture. The nodes become part of a P2P
according to the cooperating group IDs assigned
overlay network by becoming a part of a group with no
to each nodes and the size limit of CG shows
super peer which means on the other hand that there is
weather the nodes are idle or waiting for any
no hierarchy in the network. Formation of groups en-
other nodes in the network.
hances the cooperation by identifying the node with its
cooperating group ID (CG-ID). The node (peer) shows
The idea of grouping the nodes in the network is also
cooperation by the number of times it changes the CG-ID.
very common where the nodes are clustered into groups.
In this case a payoff value is assigned to those nodes
Mostly the grouping or clustering approaches have a
which show maximum cooperation. Through CG-ID it is
group or cluster head which is known to be the super
also possible to enhance the node search where nodes
peer of that group or cluster. However, there are ap‐
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 5, MAY 2011, ISSN 2151-9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 132
with the required resources can be searched more effi- approaches for the formation of clusters work under
ciently. some limited conditions and yet all of these approaches
give different ideas to search and share the desired file in
3.1 Cooperation in Heterogeneous Peers a purely hierarchical manner. Hence the content based
Heterogeneous peers (HP) as discussed earlier have very cluster formation requires a strong architecture where
diversified nature in the network. All the peers are virtu- content based file sharing could be enhanced on the basis
ally connected to other peers in the overlay network of node’s physical parameters. These physical parameters
where they are different in their operating systems, net- explain the existence of heterogeneous environment in
work connections, processors, RAM and memory. Hence P2P overlay network which needs to be handled before
the important aspect is to maintain an architecture in the start of file sharing procedure and also to share the
which peers can collaborate with each other without in- resources other than files on the basis of these physical
corporating free riding, malicious threats and overload to parameters.
the network. P2P systems consist of nodes that are able to 3.2 Cooperating Groups Formation for resource sharing
interact with each other and self-organize into network in Heterogeneous Peers
topologies with the purpose of sharing resources such as Grouping the peers together in P2P architecture is unique
contents, CPU cycles, storage and bandwidth [6]. Major concept but there has been a lot of study going on in this
application areas of these systems include distributed and dimension. Most of the work that has been done is related
scalable computing, internet service support, database to grouping the same peers together. This similarity is
systems, content distribution, communication and collab- mainly based on the same content that a peer has to share.
oration. This content is related a particular file or service and the
To support content transfer applications in heteroge- resources that are needed to be shared are normally ap‐
neous P2P networks there is a need of following services: plication based. In different studies this grouping of peers
A resource discovery service is needed to search together is termed as cluster formation. But this research
and locate resources. is based on development of cooperating groups of peers
A content distribution service is required for a re‐ with heterogeneous physical parameters in order to share
liable transfer of requested content to a set of physical resources as well as to improve the efficiency of
peers. node carrying relevant resource.
Up till now the resource discovery mechanisms de- The study on the issues likes secure clustering and
veloped are mostly for a semi-centralized P2P architec- grouping of peers in P2P networks has also elevated the
ture, where the super nodes receive the resource request importance of grouping in P2P networks. In 2007 S. Wang
from its connected nodes and transfer it to the other super and Y. Zhang presented a reliable self clustering method‐
nodes [7]. To maintain a resource discovery mechanism in ology by evaluating the level of trust between the nodes
pure P2P overlay networks with heterogeneous peers which can maintain the autonomous behavior of the
where there is no centralization authority there has not nodes by publishing partial knowledge of the resources
been a precise and efficient mechanism developed. Many [13]. Similarly in 2008 M. Amad and A. Meddahi present‐
experimentations and techniques to enhance cooperation ed an optimized flooding and clustering based approach
suggest that heterogeneity hinders efficiency, i.e. it reduc- for lookup acceleration P2P networks [14]. This approach
es the overall collaboration between the nodes [8]. is also based on enhancing the search of a peer according
to the required resource where the formation of clusters is
Clustering (grouping) the nodes together is one of the
hierarchical and message to acquire a particular resource
methods to enhance resource discovery. Bayesian statistic
is flooded in the network with nodes having certain TTL
analysis proposed a level of trust between the nodes for
values. This technique is scalable and practical as it deals
file sharing [9] by cluster formation. Here the criteria for with problem of disconnection during any session of a
changing a cluster are just based on the relevant file re- P2P communication but the overall criteria for grouping
quired by the node. Mostly clusters are formed in a hier- the peers together non‐hierarchically are that each peer
archical manner in which clusters have been developed (node) should participate equally and there is no super
by using layered approach [10], i.e. on the top of unstruc- peer who has the entire maintenance load to manage the
tured P2P a hybrid overlay network has been implement- cluster. Some cluster formation methods like Schelling
ed where a lot of overheads and extra protocols are re- model [15] for file sharing is based on social sciences
quired to manage the nodes hierarchically and the re- where nodes changes their clusters if they have not been
source sharing is content based particularly for multime- provided by the relevant resources from its neighboring
dia files. A. Yonezawa et.al [11] proposed a clustered ap- nodes. For this purpose selfless and selfish clustering is
proach over multicast overlay networks by introducing a introduced. In selfless clustering node drop the neighbor
cluster coefficient which is maximum physical number of if the neighbor is connected with the other nodes. While
hops between any two nodes. But this approach fails in selfish clustering node drop the neighbor even if it has
when the cluster size increases. no other connected node.
As far as nodes dealing in heterogeneous environment
3.3 Heterogeneous Cooperating Group-based Newscast
of P2P network, they can search a file by building an
Protocol
overlay network on the basis of node’s power [12]. This The current research is an enhancement of our previous
method does not seem to be efficient enough as the file research based on formation of clusters in P2P overlay
searching would be across the whole network. All the networks [5]. These clusters were formed on the basis of
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 5, MAY 2011, ISSN 2151-9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 133
physical parameters of the nodes, i.e. storage space, RAM, ed to it. The size of the cache is according to the network
processor speed etc, in such a way that cluster ID is as‐ size and is configurable through configuration file. In
signed to different nodes on the basis of capability levels. each cycle during the simulation of newscast protocol the
These capability levels are different metrics values as‐ nodes with minimum time‐stamp values are replaced
signed to a single physical parameter. The current re‐ with new nodes in the network, as in the real time situa‐
search enhances this formation of clusters by adding an tions the nodes randomly join or leave the network. The
environment of cooperation between the nodes in a P2P data structure associated to each node has been imple‐
overlay network. This environment is provided to each mented accordingly. To incorporate a clustered (grouped)
node by adding an idea of on‐demand peer configuration approach in a P2P overlay network along with the
over Heterogeneous Cluster‐based Newscast Protocol achievement of cooperation among the nodes the data
(HCNP). Peer cluster configuration, to achieve coopera‐ structure associated to HCGNP is shown in Fig. 2.
tion, is a procedure in which nodes join or leave a particu‐
lar cluster on the basis of its available capability level Node ID
(ACL) and demanded capability level (DCL). The values
of demanded and available capability levels decide that
which of the two nodes are going to cooperate with each Time
other. Each node maintains a cache in which it holds the Stamp
information of its neighboring nodes. In every cycle these
caches refresh themselves to maintain the freshest list of
A_Cap Node
the neighboring peers. This improvises the idea of main‐
Level
taining cooperation among the nodes in a network where
nodes continuously join or leave the network.
The basic objectives of the current research are: D_Cap
1. To implement a protocol that develops non‐ Level
hierarchical cooperating groups.
2. Enhancement of cooperation by developing coop‐ CG_ID
erating groups (CG) in P2P overlay networks in a
heterogeneous environment.
3. The idea of resource sharing is related to the shar‐
ing of files, services and physical resources, such
as RAM, free storage capacity, processor cycles, Fig. 2 Data Structure assigned to node in HCGNP
etc.
4. Nodes that become part of a cooperating group In HCGNP:
have to cooperate with the other nodes in that co‐ 1. Physical parameters to be shared among the
operating group. peers could be:
5. If a peer finds its compatible peer such that both a. Node’s free storage space
the peers share each other’s resources (storage, b. RAM
processor cycles, etc) simultaneously, Cooperating c. Processor cycles, etc.
Group ID (CG‐ID) is assigned to both the peers. 2. All the physical parameters are extracted
6. To find the cooperation level of peer by the rate of implicitly from the node in order to avoid
change of its CG‐ID false publishing of these parameters.
7. To identify the cooperating group with maximum 3. Assign available or demanded capability
cooperation. levels to each physical parameter according
8. To observe the rate of cooperation without giving to its capability.
any benefit ‘ß’ to the node. 4. A_Cap_Level: Physical parameter available
9. Categorize the nodes with maximum rate of co‐ to share with other node
operation for rewarding additional benefit. 5. D_Cap_Level: Physical parameter demanded
10. After rewarding additional benefit rate of cooper‐ by the node
ation is again observed. 6. Cooperating‐Group Identifier (CG‐ID) is as‐
11. To observe ß‐Cooperation (cooperation level signed on the basis of node’s physical pa‐
achieved after giving benefit to the nodes) be‐ rameters.
tween the nodes 7. Each cluster has nodes with different levels
12. To mark Tradeoffs between ß‐cooperation & co‐ of physical parameters.
operation according to the network capacity. 8. Physical parameters reflect the heterogene‐
13. To compare simulation results with SLACER al‐ ous nature of the network.
gorithm.
The data structures assigned to each peer is according Capability level refers to the value of level assign to a
to the newscast protocol with few modifications. In news‐ node on the basis of particular physical parameter, e.g. a
cast protocol the node has been assigned a “node‐ID” and node with free secondary storage of 40 GB (out of 110 GB
a “Time‐stamp” value which is just like TTL values asso‐ of total space) has assigned max capability level: “5”. As
ciated to any message segment. Each node maintains the discussed earlier that through node data structure we can
information of the neighboring nodes in a cache associat‐ extract capability level of the physical parameter of a
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 5, MAY 2011, ISSN 2151-9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 134
PEERSIM are to invoke a mechanism which ensures the The importance of nodal search in order to acquire
compatibility of the java source file and the configuration the information of the desired resources is one of the
file. In PEERSIM the basic protocol to generate nodes is most important aspects that have been studied by
Newscast Protocol. several researchers. In HCGNP node search proce‐
dure has also been observed as the protocol is also
4.1 Protocol Simulations required to be efficient in locating the nodes carrying
The simulations of the protocol is based on the configura- desired resources.
tion file which defines the P2P overlay networks size, the
protocols used to generate the nodes and the number of 3. Computational Complexity
cycles for which the protocol is to be simulated. The for- Another important feature which plays a very vital
mation of nodes in the network is such that at each cycle role in evaluating protocols is the computational
new nodes join the network and the older nodes leave the complexity associated to a particular protocol. The
network. This happens by refreshing the cache list associ- current research implements a protocol that has been
embedded in Newscast protocol to generate cooperat‐
ated to each node. As mentioned earlier the HCGNP has
ing nodes. Protocol like PD‐Protocol that has imple‐
been embedded in the Newscast Protocol which implicit-
mented SLACER on the top of Newscast Protocol
ly incorporates cooperating behavior in the nodes when
proves to be less efficient which has also been dis‐
the nodes are joining the network. This cooperating be- cussed in section below. The computational complexi‐
havior is based on the nodal resources and the physical ty associated to any implementation plays a vital role
parameters associated to the nodes. as it proves whether the achieved results are worth to
The simulations are also based on the benefit value be implemented in a real time environment.
“ß” assigned to the nodes according to the network ca-
pacity. The cooperating behavior is observed to be en-
hanced after the implementation of HCGNP. The cooper- 4.2 Output Generated
ation level of nodes is compared with PD-Protocol that The simulation results can be seen in the output file gen-
has implemented SLACER on the top of Newscast proto- erated by changing several parameters of the configura-
col. The level of cooperation has also been observed in the tion file. The number of CGs in the network show the
simulations when there is no benefit “ß” given to the cooperation rate of the nodes in a P2P overlay environ-
node. Moreover, the overall performance of HCGNP over
ment where each node possesses different physical re-
SLACER has also been observed. The following sub sec-
sources. On the basis of these resources the capability
tions will elaborate the findings of these observations.
level is assigned to that node. The formation CGs not only
The implementation of HCGNP is based on achieving
enhances the rate of cooperation in the network but also
a level of cooperation between the nodes and to improve
improves the nodal search in order to acquire the re-
the nodal search in order to retrieve information about
quired resource. The protocols developed earlier only
the relevant resources. The parameters on the basis of
enhanced cooperation among the nodes without consid-
which HCGNP has been evaluated are discussed below:
ering the network capacity and network heterogeneity
1. Cooperation while HCGNP has been implemented for the network in
To achieve a level of cooperation in the nodes there a heterogeneous environment. The level of cooperation
should be a criterion on which the node should be achieved while implementing HCGNP is according to the
judged, that whether it can cooperate or not. Every network capacity. The benefit rewarded to the nodes
single node participating in P2P overlay networks which show cooperating behavior should not overload
possess some resource. These resources could be a the network.
file, a service, free storage to share with other node,
RAM, processor cycles etc. As there is no centralized
4.3 Results and Comparisons
authority in P2P overlay networks the task of incor‐ Formation of CGs is one of the most important tasks,
porating cooperation among the peers depends upon which is achieved after creating an environment of coop-
the capability parameters assigned to the nodes ac‐ eration in the network. The results are accumulated by
cording to its available and demanded resources. The making changes in the configuration file. The following
rate of cooperation in the network is elevated if and values have been changed in the configuration file to ob-
only if two nodes cooperate with each other simulta‐ serve enhancement in cooperation between the nodes in
neously. This conditional communication between P2P overlay network:
the nodes in the network compels the nodes to coop‐ 1. Number of cycles to simulate HCGNP.
erate if it has to remain the part of a network. Hence 2. Network size
the percentage and probability of cooperation in the 3. Cache length associated to each node.
nodes communicating with each other is found to be 4. Degree size on the basis of which the cache would be
maximum. The cooperating behavior of the node refreshed in the next cycle of simulation.
while implementing HCGNP and SLACER has been 5. Benefit value given to the cooperating node at every
discussed in detail below. cycle according to the network capacity (x).
The observations of the simulations can be seen in the
graphs below:
2. Node Search The number of CGs at any time instance (t) with dif‐
ferent network sizes can be seen in Fig. 4, where the in‐
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 5, MAY 2011, ISSN 2151-9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 136
crease in the number of CGs according to the network 2. Similarly, the percentage of CNs in CG‐ID=46 is ap‐
size can be seen. We see that as the size of the network proximately 35% and the percentage of NCNs in the
increases the number of CGs also increases. This number same CG is approximately 65%.
of cooperating groups is taken at a specific time instance 3. Thus, it can be seen that as the percentage of CNs in a
(t) which can be termed mathematically as: CG increases, the percentage of NCNs decreases and
Number of CGs (t) = Network size/ cache length (as- vice versa.
sociated to each node) 4. In CGs with IDs 9, 32, 44, 46 and 64, it can be seen
The nodes that join the networks become part of a CG that more than 50% of nodes are NCNs of the group.
according to its available and demanded capability levels. 5. These nodes are termed as NCNs because they have
At every cycle the new nodes join the network and the not developed a cooperating pair with any node in
older nodes leave the network, hence the nodes which do the group.
not become the part of CG in one cycle can become the 6. Whenever new nodes join the network, first they
part of a CG in the next cycle with changed values of ca- compare their ACL and DCL values with these
pability levels. NCNs.
7. If NCNs are not able to develop a cooperating pair in
200
the entire cycle, they are discarded from the network.
180 8. This result is taken when no benefit is given to the
nodes in the network which shows that the average
N O . O F C o o p e ra tin g G ro u p s (C G s )
160
percentage of CNs is greater than the average per‐
140 centage of NCNs even when there is no benefit or
payoff value rewarded to the nodes.
120 9. This shows that an acceptable condition for coopera‐
tion is achieved without overloading the network.
100
80 The overall network state in terms of cooperation and
non‐cooperation among the nodes can be observed in Fig.
60 6. Without investing any network resources in rewarding
40 benefit, it can be seen that the percentage of CNs as com‐
pared to NCNs is always greater. Irrespective of network
20 size, percentage of optimized cooperation achieved is
higher than the percentage of non‐cooperation. Although
0
10002000 4000 8000 12000 16000 the number of non‐cooperating nodes in the overlay net‐
NETWORKS SIZE work is not minimum, it is worth retrieving such cooper‐
ating behavior with no impact on network capacity. Net‐
Fig. 4 Number of CGs at any instant of time (t) work capacity being the vital attribute required during
the communication phase of the nodes in the network,
90
CN should not be wasted in rewarding as an “extra benefit
N O N -C O O P E R A T IN G N O D E S IN E A C H C G
NCN value” to the CNs. But if network capacity allows such
80
implementations, it may be the best option.
70
60
% O F C O O P E R A T IN G &
CN
60 NCN
50
A N D N O N -C O O P E R A T IN G N O D E S
50
% O F C O O P E R A T IN G
40
40
30
30
20
20
10
10
0 10 20 30 40 50 60 70
CG (ID 1-66)
0
Fig. 5 % of cooperating and non-cooperating node in each CG with network 10002000 4000 8000 12000 16000
size = 4000 nodes NETWORKS SIZE
Fig. 6 % of cooperating and non-cooperating nodes in each network
After the implementation of SLACER and HCGNP (ß)
In Fig. 5 it can be seen that each CG carries number of
with the provision of rewarded benefit value at each cycle
nodes with different levels of cooperating nodes (CN) and
to the CNs, it can be seen in Fig. 7 that HCGNP (ß) per‐
non‐cooperating nodes (NCN). Here we can see that:
forms far better than SLACER. This improved behavior is
1. The percentage of CNs in CG‐ID=8 is approximately
attributed to the factor that maximum percentage of co‐
85% and the percentage of NCNs in the same CG is
operation in HCGNP (ß) is achieved within 10 cycles of
approximately 15%.
simulations. While in SLACER the maximum percentage
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 5, MAY 2011, ISSN 2151-9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 137
of cooperation is achieved after 100 cycles of simulations. tion is increased to 68%. And finally when the network
It is observed that during first four cycles, the percentage capacity is increased to 5x, the gain in cooperation is max‐
of CNs in SLACER is 9%, while the percentage of CNs in imum (i.e. 94%). Fig .8 also shows that the highest level of
HCGNP (ß) is 30%, for network size (NS) of 1000 nodes. cooperation achieved for each value of network capacity
This trend in cooperation is doubled within next two cy‐ takes not more than ten cycles to achieve cooperation.
cles in HCGNP (ß), i.e. 60% cooperation is achieved, while This also decreases the computational complexity and the
SLACER increased its level of cooperation to 11% at that overall percentage of cooperation remains approximately
stage. Hence SLACER reached at maximum level of coop‐ constant after achieving its highest value.
eration in 100 cycles. The observation adds to the fact that
the computational complexity while implementing 100
SLACER is far greater than implementing HCGNP (ß), as
90
to achieve the targeted value of cooperation wastes sever‐
al cycles of simulations in SLACER. But another very im‐ 80
% O F C O O P E R A T IN G N O D E S
portant fact is the wastage of network resources in re‐
70
warding extra benefit to the CNs in SLACER as well as in
HCGNP. It demands a thoughtful approach to find the 60
tradeoffs between the two protocols. 50
This difference in the level of cooperation between
SLACER and HCGNP (ß) is due to the simplified nature 40
of HCGNP. The node is forced to cooperate as soon as it 30
joins the network through HCGNP (ß). While in SLACER,
two protocols run one after the other, i.e. at the start 20
HCGNP(ß)-NS 1000
newscast protocol is configured to generate nodes with 10 HCGNP-NS 1000
Node ID and TTL value. After the formation of whole SLACER-NS 1000
network, PD‐protocol is implemented on each node. In 0
0 10 20 30 40 50 60 70 80 90 100
SLACER the delay in the gain of cooperation is because of
NO. OF CYCLES
bi‐level protocol handling. The rule of cooperation im‐
plemented in HCGNP (ß) is embedded in newscast proto‐
col. Therefore the gain in cooperation is very rapid as the Fig. 7 % of cooperation/cycle in HCGNP (ß), HCGNP (without benefit
value) & SLACER
node is configured using single protocol. At the time of
node generation, the available and demanded capability
levels are also assigned to the nodes along with the Node‐
ID and TTL value. Hence the protocol handling of the 100
nodes is very efficient and resourceful. 90
Now if we assume that there is no extra benefit given
to a node in HCGNP then we can see that 50% of the 80
% O F C O O P E R A T IN G N O D E S
centage of cooperating nodes, i.e. within 10 cycles where [2] Eng. K. Lua, J. Crowcroft, M. Pias, R. Sharma and S. Lim, “A
more than 90% of the nodes in the network show cooper‐ Survey and Comparison of Peer‐to Peer Overlay Network
ating behavior. schemes”, IEEE Communications Survey and Tutorial, March 2004.
SLACER and HCGNP, both give some extra ben‐ [3] Spyros Voulgaris, M´ark Jelasity, Maarten van Steen, “A
efit to the nodes in order to achieve cooperation. This Robust and Scalable Peer‐to‐Peer Gossiping Protocol”,
phenomenon seems to create an ideal situation in P2P International workshop on agents and peer‐to‐peer computing
overlay networks to eliminate free riding and enhance No2,2004 , vol. 2872, pp. 47‐58, 2004.
cooperation among the nodes. But realistically, while giv‐ [4] D. Hales, S. Arteconi and O. Babaoglu, “SLACER:Randomness
ing some extra benefit to the nodes the network becomes to Cooperation in Peer‐to‐Peer Networks”, Department of
overloaded. This load is obviously the extra share of Computer Science, University of Blogna, Italy, 2005 IEEE.
bandwidth provided to the cooperating nodes in form of
[5] Irum Kazmi, Saira Asalm, M. Y. Javed, “Cluster‐based Peers
a benefit value. The observations discussed in the previ‐
Configuration Using HCNP in Peer‐to‐Peer Overlay
ous chapter show that, to enhance the behavior of coop‐ Networks”, CICSyn2010.
eration in a P2P overlay network the nodes have to main‐
[6] T. K Madson, Q. Zang, F. Fitzek, M. Katz, “Exploiting
tain some capability level. This capability level should be
Cooperation for Performance Enhancements and High Data
such that it drives the node to gain its desired resource,
Rates”, Journal of Communications, VOL. 4, NO. 3, April 2009.
without creating any congestion in the network. The re‐
sults also show that the level of cooperation among the [7] Victor O. K. Li, Li Cui, Q. Liu, G. H. Yang, Z. Zhao, “A
nodes in any network depends upon the network capaci‐ Heterogeneous Peer‐to‐Peer Network Testbed”, ICUFN, IEEE,
ty for support. Hence this parameter proves to be the 2009.
most important, as the desired values of cooperation are [8] S. Schosser, K. Bohm, B. Vogt, “Competition and Cooperation
totally dependent on it. However the minimum level of in Heterogeneous Structured P2P Systems‐Are they Mutually
cooperation achieved after implementing HCGNP is also Exclusive”, University of Karlsruhe, Germany, 2007.
more than 50% in which the overall probability of the [9] S. Wang , Y. Zhang , “Reliable Self‐Clustering P2P Overlay
nodes to cooperate in the network is highest. Hence we Networks”, Department of Computer Science and Technology,
can say that HCGNP proves to be more suitable protocol Tongi University, China, COMPSAC‐IEEE‐2007.
that elevates the percentage of cooperation among the [10] Mouna Kacimi, Kokou Yétongnon, Yinghua Ma, Richard
nodes in a P2P overlay environment without over bur‐ Chbeir, “HON‐P2P: A Cluster‐based Hybrid Overlay Network
dening the network and without wasting network capaci‐ for Multimedia Object Management”, ICPADS’2005, University
ty. Thus the major advantage of cooperating‐groups in a of Bourgogne, 2005.
heterogeneous P2P overlay network is that nodes having
[11] Akinori Yonezawa, Khaled Ragab, “A Self‐organized
varying resources can cooperate well if designed on the
Clustering‐based Overlay Network for Application Level
basis of heterogeneity. Quality of service (QoS) like securi‐ Multicast”, Journal of Networks, April 2009.
ty protocols, multicasting and network management pro‐
[12] James Z. Wang, Matti A. Vanninen, ʺA Novel Self‐
tocols like tracking the current state of a network, can also
Configuration Mechanism for Heterogeneous P2P Networks,ʺ
be implemented using heterogeneous capabilities of
Intelligent Agent Technology, IEEE / WIC / ACM International
nodes.
Conference on, pp. 281‐287, 2004 IEEE/WIC/ACM International
The benefit value rewarded to the nodes is the
Conference on Intelligent Agent Technology (IATʹ04), 2004.
extra bandwidth share. But the type of benefit value may
vary according to the network size, type of the network [13] S. Wang , Y. Zhang , “Reliable Self‐Clustering P2P Overlay
and type of the network resources shared. The criteria to Networks”, Department of Computer Science and Technology,
reward a benefit to the node not only depend upon the Tongi University, China, COMPSAC‐IEEE‐2007.
network capacity, but also depend upon the nodal capaci‐ [14] M. Amad, A. Meddahi, “DV‐Flood”: An Optimized Flooding
ty and nodal state. The most important point needs to be and Clustering based Approach for Lookup Acceleration in P2P
highlighted is the maintenance of network and node state. networks”, GET/Telecom Lille 1, France, 2008‐IEEE.
HCGNP requires an application level enhancement of [15] Atul Singh, Mads Haahr, “Decentralizing clustering in pure
certain features like a well maintained authentication pro‐ P2P overlay networks using Schelling’s model”, ICC’2007, pp.
tocol is required which authenticates the nodes before 1860‐1866,2007.
they join the network. Another very important enhance‐
ment required in HCGNP is the improvement in its dy‐
namic nature in heterogeneous environment. This can be
done by expanding the concept of multiple physical pa‐
rameters under the consideration of different network
connections and different network types.
REFERENCES
[1] R. Subramanin, B. D. Goodman, “Peer to Peer computing: The
evolution of a Disruptive Technology”, Idea Group Publishing,
2005.