‫به نام خدا‬
‫‪Path Set Selection in‬‬
‫‪Mobile Ad Hoc Networks‬‬
‫سيد عليرضا كارداني‬
‫مجتبي اميرخاني‬
‫‪82256202‬‬
‫‪82206906‬‬
‫زمستان ‪1382‬‬
Contents
•
•
•
•
Introduction
Routing with Quality of Service
Single path versus Multi path
DPSP protocols
Introduction
• The nature of ad hoc network involves many
different constraints occurring in different layer of
the network (MAC layer, network layer, etc.)
• Different problems can be met :
–
–
–
–
–
–
Unpredictable link properties
Node mobility
Limited battery life
Hidden and Exposed terminal problem
Route maintenance
Security
Routing with Quality of Service
• Quality of Service (QoS) :
« QoS is usually defined as a set of service requirements
that need to be met by the network while transporting a
packet stream from a source to its destination »
Routing with Quality of Service
• Different metrics to consider QoS :
–
–
–
–
–
Delay
Bandwidth
Probability of packet lost
Delay variance
Etc.
• All applications don’t need the same QoS!
Single path versus Multi path
• Why should we use Multi-Path protocols?
– Several paths can collectively ensure the QoS
– To use at best the network resources
– To improve network reliability
• Which set of paths must be selected?
– Shortest paths (simple solution)
– DPSP (Disjoint Pathset Selection Protocol)
A simple solution
• shortest paths are found one after the other,
and are removed from graph.
• This solution does not work well in practice
• we wish to find as many disjoint paths as possible that are as reliable
as possible
An example
Simple algorithm find only one path
We can find two disjoint path
which are more reliable
DPSP : presentation
• The goal of this protocol : to propose a set of
highly reliable disjoint paths (D)
• To perform this goal, it uses a path reliability
metric which guides the route selection process
• Each node continuously monitors the reliability of
each of its incident links
• This approach is independent of the routing
protocol
DPSP - Path Selection
• The network is modeled as probabilistic
graph Gp = (V, E) where each vertex, edge
or arc has a weight p, its operating
probability (and therefore, its failure
probability (q = 1 – p) )
• For a source node s and a destination t, t  s,
Rel s  t (GP ) denotes the probability that
there exists at least one path connecting s
and t over GP
DPSP - Path Selection
• Four steps :
1. A search for the more reliable path between s and t
2. A decision to know if this path improves the path set
reliability or no
3. A means of adding path to the path set
4. An update of the basic graph
• These steps are followed until no new path
can be discovered
Path selection - first step
• Forward arc : edge that belongs to one path of D
and ‘follows’ the direction from s to t
• Backward arc : edge that belongs to one path of D
and ‘follows’ the direction from t to s
• A shortest path algorithm (SP) is use to find the
different paths
• The backward arcs can temporarily be used to
define a path. Hence, interlacing can occur.
Path selection - definition
• a weight
is assigned to each
arc.
• for an s  t path Pi, with (vi, vj) the arcs that
belong to Pi,
• The path set operate with probability define
as:
Path selection - definition (cont)
• We also define the constant
• We used this constant to change the weight of arcs.
Path selection - second step
• Two metrics are provided to determinate
when it is suitable to include a new path or
not
• They compute the reliability of the path set
before and after the new path is added
• The reliability of a path can be seen as the
product of its different links reliabilities
Path selection - third step
• If the new path is not interlacing with another
path, the protocol simply adds it to the set
(without metrics computation)
• If the overall reliability of the path set is not
improved, it drops the path and the weights of
the links are changed in such way that they
can not be used by SP after this
Path selection - fourth step
• When a path is added, all forward arcs weights are
changed to a high positive value (Cp) and backward
arcs weights to a negative value (Cn) (so SP can use
them to find new interlacing path)
• If the interlacing is removed, the weights of the
backward arcs that belong to the interlacing are
changed back to their original value
An Example of DPSP Operation
• we assume only three different link reliability
values:
– pe = 0.9 for links shown as thick dashed lines,
– pe = 0.95 (thick solid lines) and
– pe = 0.7 for the normal lines.
Operations for example
•
•
•
•
•
First path find : P1 = {1,2,4,5,8,9,11}
The second path : P2 = {1,4,8,11} . No interlaces between P1 , P2
The third path : P3 = {1,3,7,8,5,4,6,10,11} . Interlaces with P1
{(8,5),(5,4)} and would result npath1 ={1,2,4,6,10,11} and npath2
={1,3,7,8,9,11} . metric1 > metric2 (actual values: 0.7632 > 0.5233) and
reject.
The fourth path : P4 = {1,3,7,8,4,6,10,11} . Interlaces with P2 {(8,4)} and
would result npath1 ={1, 4,6,10,11} and npath2 ={1,3,7,8,11} . metric1 >
metric2 (actual values: 0.7577 > 0.5221) and reject.
Two disjoint path find {1,2,4,5,8,9,11} , {1,4,8,11}
Operations for example
• If P(4,8) = 0.5 then three disjoint path find
{1,2,4,5,8,9,11} , {1,4,6,10,11} and {1,3,7,8,11}.
DPSP OPTIMIZATION
• worst-case computational complexity, is still
polynomial
• Optimization :
–
If the improvement in the reliability of the path set
drops below a given threshold, the method can
conclude.
– The threshold value can be a design parameter
regulating a trade-off between path set reliability and
computational overhead.
EVALUATION
• generate random multihop topologies for 100 nodes
distributed in an area of 1000 m. x 1000 m.
• assign random reliability values to links from a
normal distribution with mean pe ={0.25, 0.5, 0.85,
0.97} and standard deviation 0.0707
• For each setting, 1000 Monte Carlo iterations are
performed.
References
• P. Papadimitratos, Z. J. Haas, E.G. Sirier,
«Path Set Selection in mobile ad hoc network»
in Proc. MOBIHOC’02 Copyright 2002 ACM
• P. Mohapatra, J. Li and C. Gui,
«QoS in mobile ad hoc network»
IEEE Wireless Communicatons 10 (2003)