RFXCOM RFXtrx433 Spezifikationen PDF herunterladen (Seite 73)

Chapter 5. Evaluation 66

The following sections describe the benchmark results for each node in more detail. Each publisher,

subscriber and service was called at least 100 times and the measured processing times were

recorded and averaged.

Gateway Driver: The gateway driver was tested with valid and faulty data from the simulation

node and the WiﬁCtrl gateway. The processing times for the “protocol in” topic were too

low for being measured, whereas the “protocol out” topic needed up to 3 ms for execution,

depending on the fragmentation of the input stream. The memory usage was 7562 kB for

all states of the node, due to automatic memory allocation of the Qt framework.

Device Manager: A random database was created to test the device manager node. This

database contained 50, 500 or 5000 devices, of which half were Intertechno (IT) and the

other half HomeEasy (HE). Every fourth receiver was connected to two transmitters and

every fourth transmitter was connected to two receivers. Every tenth connection was mixed

up with both IT and HE devices to force a translation event. The measured times of all

services and topics were below the resolution of the QTime class for all three database sizes.

The memory usage was ﬁxed at 7500 kB due to automatic memory allocation of the Qt

framework.

Visualization: As the visualization node uses a local copy of the artiﬁcial database of the device

manager, the diﬀerent database sizes aﬀect the performance of the node. In order to test

these eﬀects, all database entries were divided among random scenes. These scenes are

composed of up to ten devices and their total number equals a tenth of the number of

devices of the database. Nevertheless, when using the 5000 entry database with its 500

scenes, the processing time for all services and topics is below 1 ms. Only the time of the

composition of the visualization and scene menu are that long that they can be measured.

The results are summarized in table 5.1.

Database size 50 500 5000

Visualization Menu loading time [ms] 16 121 1215

Scene Menu loading time [ms] 230 230 230

Scene Menu loading time (no background) [ms] 35 35 35

Memory Usage [kB] 26532 29376 59772

Table 5.1.: The averaged benchmark results of the visualization node.

The previous performance benchmarks showed that the performance of each node is high. As a

result, the latency of the system depends mainly on the time needed for communicating via ROS.

These timings are listed in ﬁgure 5.2 on the following page. The overall latency is slow compared

to the duration of a protocol message. This makes it possible to extend the system with additional

nodes.

1 2 ... 68 69 70 71 72 73 74 75 76 77 78 ... 94 95

Keine Kommentare

RFXCOM RFXtrx433 Spezifikationen Seite 73