ROUTING 125
BFTR follows these steps. In the route discovery, BFTR uses the DSR (Dynamic Source Routing )
[48] flooding concept to retrieve a set of paths Ω between source and destination nodes, whenever
necessary. Then BFTR selects the shortest path π from this set Ω to send packets. If within a certain
observation window, this path is identified as feasible, the algorithm continues to send packet along
it. The window keeps shifting during the packet sending to observe the most recent behavior of the
same path. Anytime it is identified as an infeasible path, the algorithm discards the path /route and
selects the next shortest path to repeat the same procedure. The summary of the BFTR algorithm
is as follows:
Route discovery
: BFTR retrieves a set of paths Ω.
Route selection
: Upon retrieving a set of paths via route discovery, BFTR selects a path π
to route packets along the shortest path from the path set Ω. Using the test heuristic over a
certain observation window, BFTR identifies and determines whether this path is infeasible
and should be rejected.
Route maintenance
: If the current path is rejected, it chooses the second shortest path to re-
peat this procedure. If all paths in Ω have been rejected, a new route discovery procedure is
triggered to retrieve a new set of paths, i.e., to renew Ω. The testing procedure then restarts
every time Ω is renewed.
5.3.2 Finding Feasible Paths
The major issue within the BFTR algorithm is to consider the problem of finding feasible paths. We
present a testing heuristic approach to maximally maintain routing end-to-end QoS performance.
The main criterion of selecting a path is its feasibility with respect to requested QoS performance,
e.g., whether a path is able to transmit packets correctly with high delivery ratio. Any path that
follows this pattern can be regarded as a feasible path and a path which deviates from this pat-
tern is an infeasible path. On the basis of this idea, data packets transmission is used as a tool to
measure the end-to-end QoS performance. By comparing the measured path performance and a
priori feasible path model, the testing heuristic determines the feasibility of a path. The presented
testing heuristic is based on end-to-end performance, thus requiring no additional support from
the intermediate nodes, which might be misbehaving. Also, from an end-to-end point of view, the
BFTR algorithm can provide a unified solution for many types of misbehavior. Before we discuss
the testing procedure, to determine whether a path π is feasible, we introduce “a priori feasible path
model.”
Feasible path model. Given a path π, we define a random variable Y (
π
) as follows:
d D
Y
1
if correctly delivers
the observed data with delay ( )
( )
0 otherwise
π
π
π
£
ì
=
í
î
1.
2.
3.
i
f
π
correctly delivers the observed data with delay d(
π
) D;
otherwise.

Get Quality of Service in Wireless Networks Over Unlicensed Spectrum now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.