Starting with this section, a new type of test designed to utilize timed obsolescence was used to produce the results. A more complex testing application than that described previously in chapter 6 was necessary to effectively simulate a real-time, 30 second, uncompressed video stream. Instead of continuously sending 10,240 byte parcels as in the prior cases, this test paces itself. At the beginning of every second, 10 parcels are sent to the XUDP server; these parcels are 14,400 bytes long, corresponding to 160x120, 6-bit grey scale images. For XUDP, the first parcel in every second is labeled as RELIABLE and every subsequent parcel gets a time to obsolescence of two seconds.
The tests are also each conducted using TCP as the transport layer, so that the impact of timed obsolescence can be properly evaluated.
Table 8.5 shows the statistical results reported by the XUDP server (top) and the receiving-side application (bottom.) The frame rate at the receiver (9.8fps) nearly matched the theoretical sending frame rate (10fps) and XUDP didn't need to resort to timed obsolescence as no parcels were dropped.
| Measurement | Result |
| Total Data Acknowledged | 4391.134 KBytes |
| Total Transmission Time | 29.779 seconds |
| Network Bandwidth Utilization | 147.457 KBytes/second |
| Average Round Trip Time | 13.8 milliseconds |
| Average Window Size | 10.2 packets |
| Total Parcels Received | 291 parcels |
| Total Transmission Time | 29.683 seconds |
| Total Parcels Skipped | 0 parcels |
| Network Bandwidth Utilization (14.4 KBytes/parcel) | 141.172 KBytes/second |
| Average Parcel Reception Frequency | 9.804 parcels/second |
Figure 8.11 is slightly different than the graph of bytes acknowledged in previous sections. This graph is recorded in the application layer by simply writing the parcel sequence number of the just received parcel and the time at which it was received to a log file. The peculiar stair-stepping is due to the sending side sending the 10 parcels at the beginning of every second, rather than evenly spacing them.
An interesting RTT distribution can be found in figure 8.13. Although instances of high RTT's occur, most of the hits fall just under 10ms.
The congestion window sizing manifested as the number of packets in the network in figure 8.14 is not a pretty sight. A large percentage of the time the window is either 1 or even zero packets in length. Future study will hopefully deal with this situation.
This brief section covers the TCP version of the previous test. TCP doesn't implement timed obsolescence, so all the parcels sent to TCP will be delivered.
| Measurement | Result |
| Total Parcels Received | 291 parcels |
| Total Transmission Time | 32.244 seconds |
| Total Parcels Skipped | 0 parcels |
| Network Bandwidth Utilization (14.4 KBytes/parcel) | 129.959 KBytes/second |
| Average Parcel Reception Frequency | 9.025 parcels/second |
Table 8.6 shows that TCP finished two seconds later than XUDP, that the throughput was lower and only 9fps was realized at the receiver. Figure 8.15 shows the graph of parcels received over time.
A comparison of the results for XUDP and TCP is displayed in figure 8.16 that shows XUDP edging ahead of TCP, XUDP ending on time (30 seconds) and TCP falling two seconds behind.
