A Simulation-based Performance Analysis of Multicast Routing in Mobile Ad hoc Networks.

Full text

Page 1

A Simulation-based Performance Analysis of Multicast Routing in Mobile Ad hoc Networks
Natarajan Meghanathan

A Simulation-based Performance Analysis of Multicast Routing in Mobile
Ad hoc Networks

Natarajan Meghanathan
Jackson State University, Jackson, MS, USA
E-mail: natarajan.meghanathan@jsums.edu
doi: 10.4156/ijipm.vol1.issue1.1

Abstract
In an earlier work, we had proposed an algorithm, referred to as OptTreeTrans, to determine a
sequence of long-living stable multicast trees for mobile ad hoc networks (MANETs) such that the
number of tree transitions is the theoretical minimum. In this paper, we study the performance of
representatives from three different classes of on-demand source-tree based distributed multicast
routing protocols vis-à-vis the centralized OptTreeTrans algorithm with respect to metrics such as the
multicast tree lifetime, number of links per multicast tree and the hop count per source-receiver path.
Appropriately, the distributed routing protocols considered are the minimum-links based Bandwidth-
Efficient Multicast Routing Protocol (BEMRP), minimum hop-based Multicast Ad Hoc On-Demand
Distance Vector (MAODV) routing protocol and the stability-oriented Associativity-Based Ad hoc
Multicast (ABAM) routing protocol. Simulation results reveal a tradeoff between the three metrics: we
observe that the three classes of distributed multicast routing protocols discover trees that have
significantly smaller lifetime than those discovered using OptTreeTrans; on the other hand, the number
of links per multicast tree and the hop count per source-receiver path incurred by the stable mobile
multicast trees of OptTreeTrans are relatively larger than those discovered by the three classes of
distributed multicast routing protocols.

Keywords: Multicast, Bandwidth efficiency, Stability of trees, Hop count, Mobile ad hoc networks

1. Introduction

A mobile ad hoc network (MANET) is a dynamic distributed system of arbitrarily moving wireless
devices with limited battery power. The wireless network bandwidth is limited, shared and prone to
interference. MANET routes are often multi-hop in nature due to the limited transmission range of the
wireless devices. In the presence of node mobility, routes between nodes frequently change and need to
be reconfigured. As a result, on-demand route discovery (discovering a route only when required) is
often preferred over periodic route discovery and maintenance, which would involve frequent
exchange of control information among nodes [2]. We restrict ourselves to on-demand routing
protocols in this paper.
Multicasting is the process of sending a single stream of data from one node to multiple recipients
by establishing a routing tree, which is an acyclic connected sub graph containing all the nodes in the
tree. While propagating down the tree, data is duplicated only when necessary. This is more
advantageous than independent unicast transmissions, from the sender to each receiver, which may
lead to network clogging. Multicasting in ad hoc wireless networks has numerous applications in
collaborative and distributed computing like civilian operations (audio/ video conferencing, corporate
communications, distance learning, outdoor entertainment activities), emergency search-and-rescue,
law enforcement and warfare situations, where establishing and maintaining a communication
infrastructure may be expensive or difficult. A common feature among all these applications is one-to-
many and many-to-many communications among the participants [20].
Several multicast routing protocols have been proposed for ad hoc wireless networks [16]. They are
mainly classified as tree-based and mesh-based. In tree-based multicast protocols, only one route exists
between a source and a destination and hence these protocols are efficient in terms of the number of
link transmissions. The tree-based multicast protocols can be further divided into two types: source
tree-based and shared-tree based. In source tree-based multicast protocols, the tree is rooted at the
source, whereas in shared-tree based multicast protocols, a single tree is shared by all the sources
4

Page 2

International Journal of Information Processing and Management
Volume 1, Number 1, July 2010

within the multicast group and is rooted at a node often referred to as the core node. Even though
shared-tree based multicast protocols are more scalable with respect to the number of sources, these
protocols suffer under a single point of failure, the core node. On the other hand, source tree-based
protocols are more efficient in terms of traffic distribution. In mesh-based multicast protocols, there are
multiple routes between a source-destination pair. Even though this could provide robustness, these
protocols do not efficiently use the network bandwidth and are not efficient in terms of the number of
link transmissions. In this paper, we restrict ourselves to studying the on-demand source tree-based
multicast protocols.
Within the class of on-demand source tree-based routing protocols, three categories of multicast
routing protocols have been identified: One category of multicast protocols called bandwidth-efficient
protocols attempt to minimize the total number of links in the tree; the second category of protocols
called minimum-hop based protocols attempt to minimize the number of hops in the paths from the
source to every receiver and the third of category of protocols attempt to find stable trees. In this
research, we study the classical Bandwidth-Efficient Multicast Routing Protocol (BEMRP) [17], the
Multicast Ad Hoc On-Demand Distance Vector (MAODV) routing protocol [19] and the Associativity-
Based Ad hoc Multicast (ABAM) routing protocol [20] respectively as representatives of the
bandwidth-efficient, minimum-hop based and stability-oriented multicast routing protocol categories.
Stability of paths and trees is an important design criterion to be considered while developing
MANET multicast routing protocols. A tree is considered to be broken even if one link in the tree is
broken or failed. Link failures in ad hoc networks mainly occur when the constituent nodes of the link
move away. Frequent attempts to discover a multicast tree could congest the network and also drain out
the battery power at the critical nodes. The battery charge available at the nodes needs to be treated
preciously as the nodes may be deployed in environments where recharging might be next to
impossible. Link failures could also thus occur when at least one of the two constituent nodes of the
link ran out of battery power. Stability of the multicast trees is essential from a Quality-of-Service
point of view too. For multi-media applications that require packets to be delivered in-order with
minimum jitter, frequent changes in the routes traversed by the packets may result in out-of-order
delivery with high jitter. For reliable data transfer applications, out-of-order delivery at the receiver
may trigger frequent timeouts leading to unnecessary retransmissions at the source side. Eventually, the
application layer at the receiver gets overloaded in handling out-of-order, lost and duplicate packets.
In an earlier work [11], we proposed a polynomial-time optimal algorithm called OptTreeTrans to
determine a sequence of stable multicast Steiner trees, referred to as Stable Mobile Multicast Tree,
such that the number of tree transitions is the theoretical minimum. Given the complete knowledge of
the future topology changes, algorithm OptTreeTrans operates based on the following greedy principle:
Whenever a multicast Steiner tree is required at a time instant t, choose the longest living Steiner tree
from t. The above strategy is repeated over the duration of the multicast session. The sequence of such
longest living stable multicast Steiner trees is called the Stable Mobile Multicast Tree. Note that in this
paper, we use the terms „path‟ and „route‟, „edge‟ and „link‟ interchangeably. They mean the same.
The high-level contribution of this paper is a simulation-based performance comparison study of the
theoretically optimal algorithm OptTreeTrans with that of the distributed multicast ad hoc routing
protocols such as BEMRP, MAODV and ABAM, representing three different categories of MANET
on-demand source-tree based multicast routing protocols. This paper is an extension of our earlier work
comparing the source tree-based MANET routing protocols with respect to the stability and link
efficiency [12]. Several earlier works (e.g., [6][8][15]) in the literature have compared the MANET
multicast routing protocols. However, we could not find any work in the literature comparing the
performance of the distributed multicast protocols with that of a theoretically optimal routing algorithm.
The rest of the paper is organized as follows: Section 2 briefly describes algorithm OptTreeTrans and
Section 3 gives brief descriptions of the BEMRP, MAODV and ABAM protocols. Section 4 describes
the simulation environment and presents the results. Section 5 concludes the paper.

2. Algorithm for the Optimal Number of Tree Transitions (OptTreeTrans)

The problem of determining the multicast Steiner tree is that given a weighted network graph G = (V,
E) where V is the set of vertices and E is the set of edges connecting these nodes, and a subset S  V of
5

Page 3

A Simulation-based Performance Analysis of Multicast Routing in Mobile Ad hoc Networks
Natarajan Meghanathan

vertices called the multicast group or Steiner points, we want to determine the set of minimum-weight
edges of G that can connect all the vertices of S and they form a tree. In this paper, we assume the
weight of each edge is unity and that all the edges of the Steiner tree are contained in the edge set of
the graph. Accordingly, we define the minimum Steiner tree as the tree with the least number of edges
required to connect all the vertices in the multicast group (i.e., the set of Steiner points). The problem
of determining a minimum Steiner tree in an undirected graph like that of the unit disk graph is NP-
complete. Efficient heuristics (e.g., [7]) have been proposed in the literature to approximate a minimum
Steiner tree.
Algorithm OptTreeTrans uses the notion of a mobile graph and mobile tree which are defined as
follows: A mobile graph [4] is defined as the sequence GM = G1G2 … GT of static graphs that represents
the network topology changes over some time scale T. In the simplest case, the mobile graph GM =
G1G2 … GT can be extended by a new instantaneous graph GT+1 to a longer sequence GM = G1G2 … GT
GT+1, where GT+1 captures a link change (either a link comes up or goes down). But such an approach
has very poor scalability. In this research work, we sample the network topology periodically for every
one second, which could, in reality, be the instants of data packet origination at the source. For
simplicity, we assume that all graphs in GM have the same vertex set (i.e., no node failures).
Given the complete knowledge of future topology changes, the algorithm operates on the following
principle: Whenever a multicast tree connecting a given source node to all the members of a multicast
group is required, choose the multicast tree that will keep the source connected to the multicast group
members for the longest time. The above strategy is repeated over the duration of the multicast session
and the sequence of stable multicast Steiner trees obtained by running this algorithm is called the
Stable Mobile Multicast Steiner Tree. We use the Kou et. al‟s [7] well-known O(|V||S|2) heuristic (|V| is
the number of nodes in the network graph and |S| is the size of the multicast group) to approximate the
minimum Steiner tree in graphs representing snapshots of the network topology. We give a brief
outline of the heuristic in Figure 1. An (s-S)-tree is defined as the multicast Steiner tree connecting a
source node s to all the members of the multicast group S, which is also the set of Steiner points. Note
that sS.


Input: An undirected graph G = (V, E)
Multicast group S  V
Output: A tree TH for the set S in G

Step 1: Construct a complete undirected weighted graph GC = (S, EC) from G and S where (vi, vj) 
EC, vi and vj are in S, and the weight of edge (vi, vj) is the length of the shortest path from vi to vj in G.
Step 2: Find the minimum weight spanning tree TC in GC (If more than one minimal spanning tree
exists, pick an arbitrary one).
Step 3: Construct the sub graph GS of G, by replacing each edge in TC with the corresponding shortest
path from G (If there is more than one shortest path between two given vertices, pick an arbitrary one).
Step 4: Find the minimal spanning tree TS in GS (If more than one minimal spanning tree exists, pick
an arbitrary one). Note that each edge in GS has weight 1.
Step 5: Construct the minimum Steiner tree TH, from TS by deleting edges in TS, if necessary, such that
all the leaves in TH are members of S.

Figure 1. Kou et. al‟s Heuristic [7] to find an Approximate Minimum Steiner Tree

Let GM = G1G2 … GT be the mobile graph generated by sampling the network topology at regular
instants t1, t2, …, tT of a multicast session. When an (s-S)-tree is required at sampling time instant ti, the
strategy is to find a mobile sub graph G(i, j) = GiGi+1… Gj such that there exists at least one
multicast (s-S)-tree in G(i, j) and none exists in G(i, j+1). A multicast (s-S)-tree in G(i, j) is selected
using Kou‟s heuristic. Such a tree exists in each of the static graphs Gi, Gi+1, …, Gj. If there is a tie, the
(s-S)-tree with the smallest number of constituent links is chosen. If sampling instant tj+1 ≤ tT, the above
procedure is repeated by finding the (s-S)-tree that can survive for the maximum amount of time since
6

Page 4

International Journal of Information Processing and Management
Volume 1, Number 1, July 2010

tj+1. A sequence of such maximum lifetime multicast Steiner (s-S) trees over the timescale of GM forms
the Stable Mobile Multicast Steiner Tree in GM. The pseudo code is given in Figure 2.


Input: GM = G1G2 … GT, source s, multicast group S
Output: (s-S)MobileStabletree // Stable-Mobile-Multicast-Steiner-Tree
Auxiliary Variables: i, j
Initialization: i=1; j=1; (s-S)MobileStabletree = Φ

Begin OptTreeTrans
1 while (i ≤ T) do
2 Find a mobile graph G(i, j) = Gi  Gi+1  …  Gj such that there exists at least one (s-S)-tree
in G(i, j) and {no (s-S)-tree exists in G(i, j+1) or j = T}
3 (s-S)MobileStabletree = (s-S)MobileStabletree U {Minimum Steiner (s-S)-tree in G(i, j) }
4 i = j + 1
5 end while
6 return (s-S)MobileStabletree
End OptTreeTrans

Figure 2. Pseudo Code for Algorithm OptTreeTrans

3. Review of On-Demand Source Tree-Based Multicast MANET Protocols

In this section, we briefly review representative protocols from three different categories of on-
demand source tree-based MANET multicast protocols: based on minimum-links per tree; based on
minimum-hop count per path from source to receiver and based on stability.

3.1 Bandwidth-Efficient Multicast Routing Protocol (BEMRP)

According to BEMRP [17], a newly joining node to the multicast group opts for the nearest
forwarding node in the existing tree, rather than choosing a minimum-hop count path from the source
of the multicast group. As a result, the number of links (edges) in the multicast tree is reduced leading
to savings in the network bandwidth. On the other hand, the number of hops in the paths from the
source to receiver increases leading to increased delay in the delivery of data packets from the source
to individual receivers.
The multicast tree construction is receiver-initiated. When a node wishes to join the multicast group
as a receiver, it initiates the flooding of Join control packets targeted towards the nodes that are
currently members of the multicast tree. On receiving the first Join control packet, the member node
waits for a certain time before sending a Reply packet. The member node sends a Reply packet on the
path, traversed by the Join control packet, with the minimum number of intermediate forwarding nodes.
The newly joining receiver node collects the Reply packets from different member nodes and would
send a Reserve packet on that path that has the minimum number of forwarding nodes from the
member node to itself.
To provide more bandwidth efficiency, the tree maintenance approach in BEMRP is hard-state
based, i.e. a member node (including the receiver) transmits control packets only after a link breaks.
BEMRP uses two schemes to recover from link failures: Broadcast-multicast scheme – the upstream
node of the broken link is responsible for finding a new route to the previous downstream node; Local-
rejoin scheme – the downstream node of the broken link tries to rejoin the multicast group by using a
limited flooding of the Join control packets.

3.2 Multicast Ad Hoc On-Demand Distance Vector (MAODV) Routing Protocol

MAODV [19] is the multicast extension of the well-known MANET unicast routing protocol: Ad
hoc On-demand Distance Vector (AODV) routing protocol [18]. Here, a receiver node joins the
7

Page 5

A Simulation-based Performance Analysis of Multicast Routing in Mobile Ad hoc Networks
Natarajan Meghanathan

multicast tree through a member node that lies on the minimum-hop path to the source. As a result, the
hop count of the paths from the source to the receivers will be low.
A potential receiver wishing to join the multicast group broadcasts a Route-Request (RREQ)
message. If a node receives the RREQ message and is not part of the multicast tree, the node
broadcasts the message in its neighborhood and also establishes the reverse path by storing the state
information consisting of the group address, requesting node id and the sender node id in a temporary
cache. If a node receiving the RREQ message is a member of the multicast tree and has not seen the
RREQ message earlier, the node waits to receive several RREQ messages and sends back a Route-
Reply (RREP) message on the shortest path to the receiver. The member node also informs in the
RREP message, the number of hops from itself to the source. The potential receiver receives several
RREP messages and it selects the member node which lies on the shortest path to the source. The
receiver node sends a Multicast Activation (MACT) message to the selected member node along the
chosen route. The member node and all the intermediate nodes in the chosen path update their
multicast table with the state information stored in the temporary cache. The route from the source to
the receiver in the multicast tree is now setup.
Tree maintenance in MAODV is based on the expanding ring search (ERS) technique using the
RREQ, RREP and MACT messages. The downstream node of a broken link is responsible for
initiating the ERS to issue a fresh RREQ for the group. This RREQ contains the hop count of the
requesting node from the source and the last known sequence number for that group. The RREQ can be
replied only by those member nodes whose recorded sequence number is greater than that indicated in
the RREQ and whose hop distance to the source is lower (to ensure that there is no loop).

3.3 Associativity-Based Ad Hoc Multicast (ABAM) Routing Protocol

ABAM [20] constructs the multicast tree based on link stability rather than hop distance. The
stability of a link with a neighbor is characterized by the associativity ticks, which is the number of
beacons periodically received from that neighbor since the link was formed. Each node stores the value
of the associativity ticks with its neighbors. The source node initiates the tree construction phase by
broadcasting a Multicast Broadcast Query (MBQ) message in the network to inform all potential
receivers. When an intermediate node receives the MBQ message, it appends its node ID, the
associativity ticks with the upstream node and then rebroadcasts it. Upon receiving several MBQ
messages through different paths, a receiver node of the multicast group selects the most stable path
and sends a MBQ-Reply packet along the selected path. After receiving MBQ-Reply packets from each
receiver of the group, the source sends MC-Setup messages to all receivers in order to establish the
multicast tree.
Tree maintenance in ABAM is by using a Local Query Reply cycle. The upstream node of the
broken link attempts to fix a route to the receiver by broadcasting a Local Query message with TTL
value of 1. When the receiver receives the Local Query message, it responds with a Local Reply
message. The upstream node then sends the MC-Setup message to the receiver. If the upstream node
could not find a route to the receiver, then it transfers the responsibility of fixing the route to its
immediate upstream node on the path from the receiver to the source. This upstream node then initiates
a broadcast of the Local Query message with TTL value of 2. This procedure is continued until the
timer at the receiver expires and it sends a Join Query message to join the multicast group.

4. Simulations

We implemented the distributed BEMRP, MAODV and ABAM multicast routing protocols in the
ns-2 (version 2.28) simulator [3] and the centralized OptTreeTrans algorithm (abbreviated as OTT in
the Performance Figures 3 through 11) in a custom-built discrete-event simulator developed by the
author in Java. This simulator has been used successfully to implement several centralized MANET
routing algorithms recently proposed (e.g., [11][13][14]) by the author. The mobile graph is generated
by sampling the network topology (obtained through the mobility trace files) for every 0.25 seconds.
We consider a square network of dimensions 800m x 800m. The transmission range of the nodes is
8