Image Processing Reference
In-Depth Information
scenarios, many factors that cannot be captured in simulation and analysis methods, can be used to
narrow in on a realistic sensor network deployment and simultaneously extract behavioral trace data
from the system under test. Such a testbed can be located locally or even be integrated remotely. Even
federated testbeds, each with specific operating environment characteristics or dedicated to a certain
project or design team are feasible. Test jobs consist of a compilation process that is handled by the
CI framework with some additional instrumentation: () job formation with subsequent submission
to a testbed, () distribution of code to target devices, () synchronous start of all target devices, and
()logilecollection.Dependingonthecontextofthetest,analysiscantakeplaceonline,e.g.,for
the monitoring of operation or in more detail off-line after completion of a test job.
11.4.5.1 Data Collection—Basic Logging
As already discussed in Section .., CI and regression testing based on changes of the software
codebase allows for gradual refinement. This allows to continuously check the status and progress
of the integrated base of a project. Using continuous testing strategies, a design point that has been
reached once and tested to satisfaction will not be lost again. In combination with the execution on a
testbed additional tasks can be performed such as the variation and tuning of parameters (sensitivity
analysis) or the variation of the network topology by using different subsets of nodes or even different
platforms in the testbed. An example is the parametrized project of a low-power gathering applica-
tions, where the effects on changes on different sleep intervals are visualized as shown in Figure ..
Without very elaborate customization or the incorporation of complicated tool setups, the inte-
gration of CI and a testbed allows to test, repetitively using a number of different analysis methods.
Through the integrated approach however, execution is greatly simplified and data from all test jobs
is logged in a repository that references the actual software code version under test. This assures a
maximum of transparency and the ability to post facto analysis as the comprehensive system context
is maintained alongside the test results. A widely neglected fact is that when testing is carried out
manually, notes are taken in a lab notebook and log files are organized in individual, nonpersistent
structures making it almost impossible to compare results from subsequent test runs. Test runs are a
previous resource and must be maintained and organized for easy reference.
Mean current consumption
1.75
1.50
1.25
1.00
0.75
0.50
7-Mar
14-Mar
21-Mar
28-Mar 4-Apr
Date time
11-Apr
18-Apr
25-Apr
2-May
telosb-lpl500
telosb-lpl1000
telosb-lpl1500
telosb-lpl2000
FIGURE .
Parametrized data collection application project: plot of average power consumption for different sleep
intervals.
 
Search WWH ::




Custom Search