In Fig. 9, each data point is the average value of 20 experiment results. In this section, we illustrate the detail of a single experiment by showing the performance of each module of the application, in order to get an idea where the error comes from.
Fig. 10 shows the experimental results for the compressed data transmission, with 6Mbps of UDP competing traffic. Each point in this figure represents one experiment. The same experiment is repeated 52 times, with the transmitted data size increased by 16KB each time. The four figures show the predicted and measured values for total execution time, compression time, sending time, and data size after compression. Circle points are predicted value, and star points are measured value. The prediction for compression time is very accurate (see the second and third figure). It shows big difference only in two cases, which can be explained by temporary host system perturbations. Prediction for data transmission time shows a somewhat large difference with the measured value, we will discuss it in Section 3.6. But the overall error is not significant, as shown in the first figure in Fig. 10. This figure shows that the application prediction error mainly comes from the data transmission part.