Surya Bahadur Kathayat
Information and Communication Technology(ICT) Program, Asian Institute of Technology (AIT), Thailand
Nandana Rajatheva
Telecommunication(TC) Program, Asian Institute of Technology (AIT), Thailand
Keywords: Application Layer Multicasting, e-Learning, Group Management, Peer-to-Peer (P2P) Technologies.
Abstract: Traditional client-server based e-Learning architecture has many limitations. There is the overhead in a
single learning server and inefficient use of resources. There is a lack of real-time interactive ness among
learning group members therefore learning is not effective. There is also difficulty in the collaborative work
because learners may have different interest, may feel lonely and may leave the system as well.
Synchronous, any time and any where, and interactive e-Learning architecture where every learner in the
learning group can contribute their resources like in traditional class-room based system, is the requirement
of next generation e-Learning architecture. In this paper we purpose novel mechanisms for next generation
e-Learning architecture using alternative technologies, peer-to-peer technologies. The proposed framework
is based on P2P architecture for scalability, robustness, efficient sharing of resources and interactivity.
Purposed system also incorporate efficient and reliable interest based e-Learning grouping and management
mechanisms in the top of application layer. Such Interest based interactive P2P based group management
mechanisms for e-Learning will combine the tools that are already available, independent of the installed
infrastructure and offer a great deal of potential for workgroup collaboration, communities of practice, and
self-directed learning.
Various forms of e-Learning that have been
deployed so far are based on client-server
technology where learning management server plays
a key role providing contents, connectivity and
services to the members of the learning system, as
shown in figure 1. Though client-server system in
the e-Learning is easy to implement and cost
beneficial there is wastage of resources in the
system, less interactive and collaborative besides the
possibility of single point overhead and failure.
Peer-to-Peer (Nowell et al, 2002), though is not a
new technology, however only recently has been
exploited throughout the Music and entertainment
industry especially sharing content files (Lee et al.,
2002) containing audio, video, data or anything in
digital format, and real-time data. As the peers in the
Peer-to-Peer computer network relies on the
computing power and bandwidth of the participants
in the network rather than concentrating these in a
relatively few servers, P2P technology will also help
many of the limitations of the traditional e-learning
Basic motive of this work is that P2P, in concept,
can also be a natural tool for educators allowing the
Figure 1: Traditional e-Learning Model.
Bahadur Kathayat S. and Rajatheva N. (2006).
In Proceedings of WEBIST 2006 - Second International Conference on Web Information Systems and Technologies - Society, e-Business and
e-Government / e-Learning, pages 339-346
DOI: 10.5220/0001241603390346
group collaboration, management and sharing of
resources for a constructivist approach to the
learning. One scenario of such motive is briefly
explained as below. In the traditional e-Learning
system, single learning management server is
responsible for handling large number of users
which limits the problem of scalability, overhead,
and inefficient use of resources. Besides that the
interest of the different learner may be different and
so that it is very difficult to do a collaborative work.
If there is lack of collaboration then a user may feel
lonely and there is chance of leaving the system and
hence the effectiveness of the learning system will
be significantly reduced. These limitations are
inevitable even if it is assumed that if there is only a
single common interest group, like learners of a
single class room, in the client-server based e-
Learning system. Therefore it is interesting to group
the users of e-Learning system according to their
interest and apply decentralized P2P technology
within and among such groups. This approach will
result better resource utilization because every peer
can contribute their resources and more interactive
ness because each peer can communicate in two
ways either group mode via overlay multicasting or
peer mode with other peer.
There are many users in the realistic large
learning domain and we assume that these users
represent the node or peer in the overlay network
(Zhang and Hu, 2003). Different peers may have
different interest so peers having common interest
will be organized together to form a group. There
will be two possibilities; either the peer may join the
already existing group or peer may create its own
group and other peers may join it later. So there will
be two categories of the users in the group creators
or leaders and the normal users. For simplicity, if we
assume that learning will be done by chatting (not
limited to this) then interactive e-learning scenario in
such particular case will be as follows:
First the creator, say c1, will create a group
according to its interest, say computer network, and
seeking for the other interested peers. If other peer,
say p1, in the overlay network also have the same
interest in the computer network, it will first find out
the creator ‘c1’ and then pop-up chat window will
appear for the learning by chatting. Similarly if other
peers having same interest may find the group and
join the group learning process. As shown in the
Figure 2, there will be many possible cases. Most
likely case is that one peer may have more than one
interest and would like to participate in on the
multiple groups. For example peer ‘p1’ may have
common interest in the computer network and it is
already the member of the group ‘g1’, peer p1 may
also have interest in the database design so it may
wish to join the database group, say g2, (at the same
time) and get involved in the learning process.
Another likely case will be that there will be
more than one groups having nearly common
interest and either members or groups leaders may
wish to merge these groups. It is also interesting to
consider peer in a particular group may leave the
group and or may wish to create another group
having different interest than that of the current
group and advertise its group members.
Briefly, framework of P2P groups’ management
mechanisms (interest based group formation,
efficient group communication and groups
management) will be proposed (potential use in
collaborative learning) to incorporate interactive
ness among the members, allow the efficient use of
resources reducing the overhead in the server and
single point of failure, and add scalability,
decentralization and many more.
Rest of the paper is organized as below. Section
2 describes the statement of the problem and section
3 discusses about the objective and scope of this
work. Section 4 explains briefly about the related
literature on the P2P technology and grouping
mechanisms. Finally section 5 of this paper
discusses about the proposed system. Peer-to-Peer
interest based grouping mechanism, efficient data
delivery mechanism and management mechanisms,
and learning environment model are also included in
last section.
Existing Client-Server (C/S) based e-Learning
systems are facing many problems like inefficient
use of resources, single point overhead and failure,
limited interaction among the members, scalability
etc. With these limitations, C/S based e-Learning
could not be significant alternative to the traditional
classroom-based learning. P2P technology, which is
a hot technology recently for the online music and
file sharing, has potential applications on the e-
Learning as well. However, to date, there is very
limited use of P2P in the e-Learning. From
instructors’ point of view, it is challenging and
interesting to create interest based group, sub-group
formation, and merging groups having similar
interest. From the students’ point of view, the
challenge is to join into the group having specific
interest and to get the multiple group membership.
Common challenge for both is to efficiently
distribute messages to other members.
To the best knowledge there are no existing e-
Learning models for the collaborative learning using
the structured P2P network especially for interest
based group formation, for merging of two groups
having nearly similar common interests and also
group splitting or sub-grouping if the interest among
the members of the group are in conflict. Therefore
it is interesting to the design a semi-decentralized e-
Learning framework that will provide efficient and
more effective, collaborative and synchronous
learning environment.
The major objectives of this research is to purposing
a design of the group management mechanism for
structured P2P network which will incorporate
virtual ring interest based group formation, multi-
virtual ring based data delivery mechanism, and
group merging and group splitting.
This work focuses on more fundamental issues
like peer organization, group communication and
fundamental management issues for the
collaborative synchronous e-Learning. Peer
organization includes basically group formation
based on an efficient multicast group ring; joining
nodes, leaving nodes in the group are also handled
with ring repair mechanism. Data forwarding will be
based on the multi-virtual multicast-ring (multi-
unicast and unicast based) and group leader plays an
important role for the group communication. The
potential scope of proposed mechanisms or
algorithms is that these can be suitably applied for
the synchronous, effective, and collaborative e-
Learning system.
4.1 P2p Technology
According to (Rowstronand and Druschel, 2001),
P2P is a network architecture in which nodes are
relatively equal, in the sense that each node is in
principle capable of performing each of the
functions necessary to support the network. In
(Pandurangan, 2001), P2P systems are defined as the
distributed systems without any centralized control
or hierarchical organization. The software running at
each node is equivalent in functionality.
There are three types of well-defined P2P
architectures namely pure, hybrid and hierarchical
architectures. In pure P2P architecture (Schollmeier,
2002), all functions and all relevant digital objects
are distributed across many nodes, such that no node
is critical to the network's operation and hence no
node can exercise control over the network.
Flooding and document routing algorithms are used
for peer search and resource discovery in this P2P
architecture. An example of pure P2P architecture is
original Gnutella. In hybrid architecture, index is
centralized like C/S system, therefore peers first
contact the central peer to locate other peers and
shared resources. Example of hybrid architecture is
Napster (Thilliez et al., 2003) where the index is
accessed in client-server mode, whereas the digital
objects are transferred directly among peers. In
hierarchical architecture, index is hierarchically
structured and accordingly hierarchy of normal peers
and super peers are maintained.
GRID computing architecture also seems like
P2P but there are some differences. Grid computing
is a means whereby available processing resources
can be located, used and coordinated; whereas P2P
encompasses both processing and data resources.
Grid computing also differs from P2P in that it is
largely pragmatic engineering effort, rather than
scientifically designed architecture (Zhuge, 2005).
4.2 Grouping Mechanisms in P2P
Distribute hash tables (Eastlake and Jones, 2001; Sit
and Morris, 2002) are the core for the routing in the
P2P networks. Major structured P2P protocols are
Pastry (Zhang and Hu, 2003), Tapestry (Zhao et al.,
2004), Chord (Stoica et al., 2001), CAN (Ratnasamy
et al., 2001) etc. All of them take, as input, a key
and, in response, route a message to the node
responsible for that key. The keys are strings of
digits of some length. Nodes have identifiers, taken
Figure 2: P2P Group Formations.
from the same space as the keys (i.e. same number
of digits). Each node maintains a routing table
consisting of a small subset of nodes in the system.
When a node receives a query for a key for which it
is not responsible, the node routes the query to the
neighbor node that makes the most “progress”
towards resolving the query. The notion of progress
differs from algorithm to algorithm, but in general it
is defined in terms of some distance between the
identifier of the current node and the identifier of the
queried key. Some group multicasting algorithms are
pastry based SCRIBE (Castro and
Rowstron, 2002)
which is a reverse-path forwarding tree based
publish/subscribe System, Tapestry based Bayeux
(Zhao et al., 2001) which uses forward-path
forwarding tree, and Brog (Zhang and Hu, 2003)
which uses the same concept but the multicast is
formed by the hybrid approach i.e. reverse path and
forward path tree approach. Controlled and directed
flooding concepts are also used for the mini-CAN
and CHORD multicasting (Ratnasamy et al., 2001).
In distributed environment (Plaxton et al., 1997)
group multicasting can be done either by mesh or
multicast tree or ring. Mesh strategy provides more
than one path between the group members and in
tree case a single path is established between any
pair of nodes. It is also feasible to apply a mesh first
followed by tree construction algorithm to
implement overlay multicast. Mesh provides routing
stability and QoS but Tree approach have
advantages in terms of link stress, no routing loops
and simplicity. Traditional tree approaches use root
based approaches for forwarding the messages
which is well suited for the 1-to-m multicast. If the
sender desires to send the message to the multicast
group, it sends the message to the root which in turn
forwards the message along the tree to all receivers.
Network efficiency can be improved by using a
source based tree algorithms in which each source
builds an optimal tree from the source to all
receivers in the group. However this approach
introduces more overload as each node must run the
routing algorithm and must maintain large amount of
supporting information. So there are different
alternatives for the overlay multicast protocols and
existing initiatives tends to focus on the specific
optimization parameters for the targeted application
environment. Most Tree based approaches are
proposed and implemented in the structured P2P
overlay that has lower data delivery percentage with
no back up path to each member (as in ring
topology) but provide lower path stretch or link
5.1 Technological Infrastructure
Both instructors and students in the e-Learning, like
in the class room based learning, construct their own
domain and it is at least somewhat different from
others. These are self organizing and towards
decentralization. Recent technological developments
on the self-organizing and decentralized P2P
network substrate, like Pastry (Zhang and Hu, 2003),
Tapestry (
Zhao et al., 2004), CHORD (Stoica et al.,
2001), and CAN (Ratnasamy et al., 2001), point to a
new paradigm for research and for building
distributed applications. Each of these overlays
implements a scalable, fault-tolerant distributed hash
table for node ID, object ID representation and also
for limiting the number of routing hops to locate
them. So the new platform for e-Learning will be
designed on such substrates where each instructor
and/or learner node ID (based on IP address) and
their class or group IDs (Based on Group Name) is
uniquely obtained and uniformly distributed using
the SHA-1 (Sit and Morris, 2002) hash function.
Every node is identified by m-bits on the overlay
network. All the nodes that are the members of all
the particular groups will be the nodes in the
common domain. Interest based groups like classes
and common domain is like university where there
are many mini-domains.
5.2 Peer Organization in Group
Here, Ring Based Group formation over the overlay
network is proposed, such mechanisms to the best of
knowledge, in structured P2P network, are not
proposed and implemented yet. Main limitation of
the ring based multicast group is the routing delay,
but node degree is constant and they are suitable for
secure, reliable and ordered delivery of messages,
and effective against single node failure. If the
routing delay is reduced in the ring topology, then it
will be suitable for more cases, therefore a similar
group formation mechanism (virtually multi-ring
group multicast) is proposed here. Groups are
assumed to be a medium sized classes having 10 -
100 peers and the group ID will determine or
represent the group’s interest.
5.2.1 Group Formation Mechanism
One node, having sufficient resources like
bandwidth (BW), CPU, memory etc and willing to
contribute more resources, can create a group
specifying its interest in the structured P2P substrate
and wait for other nodes. Other nodes will join
according to their own interest; the virtual ring
topology will be maintained in the overlay network.
If any node wants to join the pre-existing ring, it will
request a found first peer (bootstrap node) on the
ring and the peer on the ring replies to the requesting
node and forward the request towards the root i.e.
leader of the group. After getting an
acknowledgement from the leader, bootstrap node
will reply to the requesting node along with its
neighbor information (IP address, other existing
group information) so that requesting node can join
the ring. Then all these nodes (bootstrap node,
neighboring node and requesting node) will update
the neighbor list and the leader will update group
information. Group Leader will send periodic live
signal, root information, number of nodes in the
system. When the particular node leaves the group,
then neighbor nodes will know about it from the
regular neighbor update information and accordingly
maintain their new neighbor list and inform to the
group leader to maintain updated correct group
5.2.2 Efficient Data Delivery Mechanism
Each learning peer will contain more than one (say
N/2) successor and predecessor list, so virtually
there will be more than one ring (say N/2 rings) for
multicasting the group message. Each peer will get
N (number of nodes in the group) from the group
information circulated by the instructor (root node).
Each node will also issue special request signal
(node address, SUB-COUNT) in each direction to
maintain neighbor list. The initial setting for SUB-
COUNT value is N/2. When the ring node get that
special request signal, there will be two possibilities
at the node, (i) node will reply (node address) to the
requester if SUB-COUNT is greater than zero and
then forward that request signal by decrementing the
Figure 3: Group Ring and virtual multi-ring.
SUB-COUNT value by 1, (ii) node will discard and
terminate the special request signal if SUB-COUNT
is equal to or less than zero.
Root node can control and manage the token for
ordered data-delivery within the group. Each node
willing to send data has to send data to all of its
neighbors (say N/2) using the multi-unicast
mechanism, other peers will forward the message to
the highest successor/predecessor (formation of
multiple virtual rings). Each node in the group can
forward message to the original ring and
corresponding virtual ring. This mechanism will also
reduce the overhead in the node and routing delay
(in terms of hops) at most will be improved by N/2
times than normal ring which can be mathematically
expanded as
i. Source node can send data to
number of
nodes in once in both direction of ring.
ii. Since
nodes get message in one hop, 1
node will get message in average
(unitary method) with reference to the original
nodes (half of the nodes in symmetrical
ring) will get message in
N Hops
iv. Without multiple-virtual mechanism,
nodes in the ring will get message in
v. Therefore routing delay improved will be
improved as
As an example shown in the Figure 3, there are
16 nodes in a ring i.e. in a group. According to the
proposed data delivery mechanism, there will be
16/2 = 2 virtual-rings in the group ring. The routing
delay will be improved (optimum case) by 2 times.
Similarly if there are 100 nodes in the system, there
will be 5 virtual rings within the group ring and
routing delay will be reduced by 5 times. Besides
reducing the routing delay, concept of the virtual
ring will be useful as the backup link to the normal
ring in the case of the failure of the particular node
in the normal ring. Each node will keep the source
information and maximum sequence number of the
packet that it received from that source. Each node
will then forward the received packet if that packet
is not already received from the corresponding
source; otherwise it simply discards. Suppose at time
t, node ‘3’ get the packet ‘n’ from node ‘1’ in one
hop using virtual ring. At time t’ (t’ > t), if node ‘3
get the same packet from node ‘2’, node ‘3’ will
simply ignore it which is shown by thick line in
Figure 3.
5.2.3 Group Merging and Splitting
As the interest of the peer or the learner may change
from time to time, it should be able to participate in
different groups having corresponding interests
accordingly, so group merging and splitting have
significant importance in the e-Learning.
For the sub-grouping, a peer having different
interest than the current group first create a new
group and deliver the message to the existing group
members so that interested peers join it later. This
sub-grouping is not be limited to the existing group
members; rather other group peers having same
interest may join the newly formed group. After
negotiation between the two group leaders, leader
for the newly formed common group is selected and
that manages the groups.
For the group merging, there are different
possible cases such as (i) one particular peer may be
the member of two groups and may know that two
groups are engaging in the similar activities, it will
then inform its leaders and two leaders can
communicate and exchange the information to
merge the group (ii) one leader may be the member
of the another group and these two can share the
information to merge the group.
5.3 Implementation Model
Learners are in application layers, internet based
overlay network. Each user run the standalone
application software (P2P software developed using
Jdk1.4.2) specifying its interests. Learners may have
different interests and there may be more than one
learner in the system having common interest. A
peer first tries to find out the existing groups with its
interest and if such groups are not found it creates a
new group (we assume that group creator have
sufficient resources) and wait for other peers to join
it. Once there are two or more members in the
group, they communicate with each other and
discussion goes on (currently only messaging). Also
in different cases as mentioned in the earlier section,
two groups can merge together and be involved in
the collaborative learning.
Experiments are going on parallel in two different
scenarios. First is the deployment of these
algorithms in the Internet. Using the FreePastry-
1.4.2 structured peer-to-peer platform, algorithms
are implemented (some in implementing phase)
varying the control variable (SUBCOUNT) to
control the number virtual rings. The results shown
in Figure 4 are some results with data measured
using 10 nodes (physically in the same laboratory) in
the internet and running the developed software on
each. Software is written in Java Jdk-1.4.2 version.
These preliminary results clearly show that there is
significant reduction in the delay in MVRing case
compared to RING case i.e. amount of time to
multicast the message in the group. Now,
experiments are towards increasing the number of
nodes to 100s of numbers and physically from
different locations.
Besides that we are conducting research to
calculate the node stress on each nodes and link
stress between the nodes, fault tolerance of the
proposed data delivery mechanism to compare its
efficiency with that of existing tree based group
communication mechanisms.
Second scenario of the experiment we are
conducting is the modeling of the internet in transit-
stub topology using GT-ITM topology generator and
simulating the performance evaluation of the
proposed algorithms using Network Simulator (NS-
2). In this case the nodes in the group are chosen
randomly and hence it is obviously not necessary
that neighbor node is the nearest node in terms of
time. Results as shown in Figure 4 show the latency
profile for the scenarios first where P2P code is run
on 10 machines and RING and MVRING algorithms
are compared with exactly implemented SCRIBE.
Figure 5 shows the result of second scenario. The
average latency that each node experience from its
predecessor in the case of the multiple-virtual ring
cases is about 2000 ms, 1700ms and 1500ms for
number of nodes (n) 50, 150 and 500 respectively,
while these values are 2600ms, 2700ms and 2900ms
in case of pure ring based grouping and data delivery
From the results in Figure 4 and Figure 5, it is
shown that nodes clear that latency in case of the
MVRing is significantly improved than in the RING
case and quite better compared to SCRIBE as well.
Experiments are going on to measure node stress,
link stress, fault tolerance of systems and efficiency
of the data delivery mechanisms.
Decentralization addresses the overhead in a
particular machine and all members of the learning
group can share resources among each other.
Grouping of the learners according to their interest
in P2P technology increases interactive ness and
effective collaboration in the e-learning system. The
virtual multi-ring based data delivery mechanism for
the application layer group multicasting adds the
reliable communication among group members.
Finally, the instructors and learners having variable
interest with time can be handled by the interest
based group merging and group splitting.
50 150 500
Avg latency
Average Latency Profile
s1 s2 s3 s4 s5 s6 s7 s8
Figure 4: Average Latency profile. Figure 5: Latency Profile for n=50,150,500.
Castro, M., and Rowstron, A. (2002). SCRIBE: A large –
scale and decentralized application-level multicast
infrastructure. IEEE Journal on Selected Areas in
Communications 20(8), 1489-1499.
Eastlake, D., and Jones, P. (2001). US Secure Hash
Algorithm 1 (SHA-1). Published in RFC3174. RFC
Editor: United States
Lee, Y., Oh, C., and Park, E. K. (2002). Intelligent
knowledge discovery in peer-to-peer file sharing. In
Proceeding of 11th International Conference on
Information and knowledge management (pp 308-
315). Virginia : ACM Press.
Nowell, D., Balakrishnan, H., and Karger, D. (2002).
Analysis of the Evolution of Peer-to-Peer Systems. In
ACM Conf. on Principles of Distributed Computing
(pp 233-242). Monterey: ACM Press.
Pandurangan, G. (2001). Building Low-Diameter P2P
networks. In Proceedings of the 42nd IEEE
symposium on Foundations of Computer Science (pp
492). Washington: IEEE Computer Society.
Plaxton, C.,Rajaram, R., and Richa, A. W. (1997).
Accessing nearby copies of replicated objects in a
distributed environment. In Proceedings of the Ninth
Annual ACM Symposium on Parallel Algorithms and
Architectures (pp 311-320 ). New York: ACM Press.
Ratnasamy, S., Handley, M., Karp, R., and Shenker, S.
(2001). Application-level Multicasting using Content-
Addressable Networks. In Proceedings of the Third
International COST264 Workshop on Networked
Group Communication (pp 14-29). London: Springer-
Rowstronand, A., and Druschel, P. (2001). Pastry:
Scalable, distributed object location and routing for
large-scale peer-to-peer systems. In Proceedings of the
18th IFIP/ACM International Conference on
Distributed Systems Platforms(pp 329-350)
Heidelberg Germany.
Schollmeier, R. (2002). A Definition of Peer-to-Peer
Networking for the Classification of Peer-to-Peer
Architectures and Applications. In International
Conference on P2P Computing (pp 202-209).
Munchen: IEEE Computer Society.
Sit, E., and Morris, R. (2002). Security Considerations for
Peer-to-Peer Distributed Hash Tables. In 1st
International Workshop on Peer-to-Peer System (pp
129-138). Cambridge: Springer-Verlag.
Stoica, I., Morris, R., Karger, D., Kaashoek, M.F., and
Balakrishnan, H. (2001). Chord: A scalable peer-to-
peer lookup service for internet applications. IEEE
Journal on IEEE/ACM Transactions on Networking
(TON), 11(1), 17-32.
Thilliez. M., Delot. T., Lecomte. S., and Bennani. N.,
(2003). Hybrid Peer-To-Peer Model in Proximity
Applications. In 17 th International Conference on
Advanced Information Networking and Applications
(AINA'03) p. 306
Zhang, R., and Hu, C. (2003). Anycast in Locality Aware
Peer-to-Peer Overlay Networks. In Lecture Notes in
Computer Science, 2816, 34-46.
Zhang, R., and Hu, Y.C. (2003). Brog: A Hybrid Protocol
for Scalable Application-Level Multicast in Peer-to-
Peer Networks. In Proceedings of the 13th
international workshop on Network and operating
systems support for digital audio and video (pp 172-
179). New Work: ACM Press
Zhao, B.Y., Joseph, A.D., Katz, and R.H., Kubiatowicz, J.
(2001). Bayeux: An Architecture for Scalable and
Fault-tolerant Wide-Area Data Dissemination. In
Proceedings of the Eleventh International Workshop
on Network and Operating System Support for Digital
Audio and Video (pp 11-20). New Work: ACM Press.
Zhao, B.Y., Huang, L., Stribling, J., Rhea, S.C., Anthony,
D. Kubiatowicz, D. (2004). Tapestry: A Resilient
global-Scale Overlay for Service Deployment. IEEE
Journal on Selected Areas in Communications, 22(1),
Zhuge, H., Sun, X., Liu, J., Yao, E., and Chen. X. (2005).
A Scalable P2P Platform for the Knowledge Grid.
IEEE Transactions on Knowledge and Data
Engineering 17(12), 1721-1736.