| Abdelmajid Khelil
| creator |
Khelil, Abdelmajid
| | date |
2007-07-26
| | | | description |
186 pages
| |
Broadcasting is a major communication primitive required by many
applications and protocols in Mobile Ad Hoc Networks (MANETs). It is
frequently deployed for content distribution, service discovery or
advertisement, and sensor data dissemination. Broadcast protocols
are also a fundamental building block to realize principal
middleware functionalities such as replication, group management and
consensus. Broadcasting in MANETs has therefore been an active area
of research recently.
Most of the research conducted on broadcasting in MANETs has
primarily focused only on carefully selected application and
evaluation scenarios. Consequently, the developed broadcasting
schemes do not yield good performance for other scenarios. Different
comparative studies show that the existing broadcasting techniques
are tailored to only one class of MANETs with respect to node
density and node mobility, and are unfortunately not likely to
operate well in other classes.
Node spatial distribution is a key issue for the performance of
broadcast protocols, since it determines the connectivity of the
MANET. Our survey of potential MANET application scenarios shows a
wide range of possible node spatial distributions and node
mobilities. This leads to that a MANET generally shows a
continuously changing network connectivity over space and time.
Therefore, a generalized solution for broadcasting that accounts for
the requirements of the various applications and adapts to the
heterogeneous and evolving node spatial distribution and mobility is
a major contribution. In this thesis, we present hypergossiping, a
novel generalized broadcasting technique for MANETs. Hypergossiping
integrates two adaptive schemes and efficiently switches between
them depending on local node density.
The first scheme is adaptive gossiping, which distributes messages
within connected parts of the MANET. We adapted gossiping as
follows. First, we established an analytical model for gossiping
through adopting the SI mathematical model from the epidemiology.
Then, we used the model to adapt the gossiping forwarding
probability to local node density. As a result, we provide a simple
analytical expression that nodes use to set the appropriate
forwarding probability depending on the current number of neighbors.
Simulation results showed that adaptive gossiping efficiently
propagates messages within a network partition independent of the
node spatial distribution and node mobility in that network
partition.
The second scheme is a broadcast repetition method, which detects
partition joins using an efficient and localized heuristic and
efficiently repeats the needed broadcasts upon detection of a
partition join. Our approach is mobility-assisted since it exploits
the mobility of nodes to efficiently deliver messages in frequently
partitioned scenarios. We defined mobility metrics that simplify the
design of mobility-assisted concepts, and used some of them to
design a mobility-aware buffering strategy, which can significantly
reduce the buffer overhead of hypergossiping.
Simulation results in the standard network simulator ns-2 show that
hypergossiping outperforms all existing strategies. Hypergossiping
significantly increases the delivery ratio for a broad range of
MANETs with respect to node density, node mobility and network load
while providing high efficiency and scalability.
| | format |
application/pdf
| | 1140288 Bytes | |