Improved Recovery Management and Routing in W-Grid, a Distributed Infrastructure for Effective and Efficient Multidimensional Data Management over Wireless Ad-Hoc Sensor Networks - Data Management in Cloud, Grid and P2P Systems - page 80

simulations and in each simulation we submitted 20000 queries to the system
and then tested network robustness by turning off each node of the network one
at a time. The simulator performed the following tasks:

- Random placement of sensors in a user-defined area;
- Generation of W-Grid coordinates at sensor exploiting implicit overhearing;
- Random generation of 20000 queries;
- Turning off of nodes at the delivery of queries, as previously described.

For each simulation run we observed:

- The variation in queries APL (Average Path Length), namely the number
of hops necessary to resolve a query, between W-Grid with LL and RD and
GPSR.
- The ratio of succeeded recovery in W-Grid scenarios;

7.1 Average Path Length Comparisons

Even if the comparison appears prohibitive, since GPSR can stay very close to
the ideal routing algorithm by using physical position of nodes, W-Grid returns
very good performances, especially considering that it does not require any kind
of information about geographic position of nodes. Figures 4, 5 and 6 show that
the number of hops (APL) is similar in W-Grid and GPSR especially when LL
is applied. Besides, the flat look of the averages with respect with the number
of coordinates shows that W-Grid behavior is stable according to that variable.

%!##$ !"

!

!#" !

Fig. 4. Query APL in a Network with an Average of 12 Neighbors per Node

7.2 Recovery Failures

The second measure we evaluate is the ratio of failure recovery which is cor-
rectly performed according to the different node densities and the number of
coordinates. We simulate two different recovery strategies:

- Active recovery;
- Lazy recovery.