Table 2 demonstrates that the CIM can successfully translate payloads
from non-RDF formats into JSON- LD payloads modeled by the DELTA
ontology. Although the XML format and the OpenADR schema were tested,
additional interoperability modules can be installed to allow the CIM to
convert payloads from other DR standards with alternative formats and
models.
Experiment 2: scaling parallel requests
All requests represented in Figs. 5 and 6 are routed through the CIMs in
this experiment. The time (in seconds) necessary to acquire the findings
after transmitting these payloads, including the translation time
involved in the data exchange, is the performance statistic for this
experiment. These findings are detailed in Table 3 and illustrated in
Fig. 7. From 1 to 600, the number of concurrent requests increases by 50
thread ticks. In order to compare the results, the averaged reaction
times were converted to a logarithmic scale.
Finally, the CIM limits the number of parallel requests to 650 for
security concerns; as a result, the maximum number of parallel requests
for this experiment was set at 600.
Based on the statistics in Table 3, it appears that certain queries have
extremely comparable response times. The Iman- Davenport test [52]
was used to determine whether there are significant statistical
differences between all of these requests, both GET and POST, with and
without translation, at a confidence level of 95%, in order to
determine whether the translations cause overhead during data
transmission.
As a consequence, the test finds no statistically significant
differences between GET (json-ld), GET (xml to json-ld), POST (json-ld),
and POST (json-ld to xml), however there are differences with POST (xml
to json- ld). This difference is plainly seen in Fig. 7, where the time
required to respond to these queries is substantially longer than the
others.
Conclusion: This experiment demonstrated that sharing data across DR
systems utilizing CIMs is a quick process with a linear trend, either
when parallel requests are received, as seen in Fig. 7. It is worth
noting that when there is no translation, 600 parallel requests are
responded in less than half a second (<500 ms).
Table 3 Averaged response times for parallel requests.