Wireless Sensor Networks Testing and Validation - Networked Embedded Systems - page 244

11.2.1.1 Characteristics of WSN Test Platform

. Scale
WSN systems range from small installments with a dozen to hundreds of nodes. Scale
refers to the capability of the test platform to capture the system in operation including
sensor node instantiations and models of the communication channels in between and
the environment the system is embedded into at scale. here are two different sources for
scaling issues: technical limitations and cost. While simulation allows for tests in large
quantities, testbed installations using sensor nodes have been limited due to the consid-
erable costs. Detailed simulation suffers from substantial computational requirements,
which necessitates simulations on a higher abstraction level in order to cope with this
scaling complexity as shown in Figure ..
Validation of WSNs needs to consider widely varying test areas. In the context of scale,
thesecanbeclassiiedintolocalproblems,directneighborhood,andglobalproblems.
Global problems may further be divided into subsets according to cluster, path, tree, or
system problems.
Local problems on a single sensor node concern traditional testing of embedded sys-
tem with strict resource limitations, e.g., driver testing for hardware interfaces such
as an IC bus. With two nodes, communication on a single link can be tested. Three
nodes allow for testing communication with collisions and hidden-terminal effects as
well as a primitive topology for routing: Multihop algorithms and protocols can be
tested using a predefined topology by setting a fixed neighborhood table. For testing
effects of dense or complex topologies, route reestablishment, and effects of partitions
and node unavailability a larger number of nodes are required.
. Visibility
Visibility describes the concept of the availability of information on the internal state of
the devices and their environment at any time in the test execution. Embedded systems
typically have a very limited visibility. he bandwidth to the system is limited, because it is

Real-world operation

Testbed

Simulation

1 5 10 100 1000

# nodes

FIGURE . Simulation allows for efficient scaling but is bounded by the models and abstractions used. Testbeds
and real deployments are closer to reality but considerably more difficult to handle or simply not as accessible, therfore
scaling is limited.