To determine how XUDP and TCP fared when faced with severe network congestion, I developed a simple client/server pair to run in parallel with the test with the singular goal of generating lots of network load. The actual 20fps tests are identical to those in the preceding section.
XUDP was worse off this time, only receiving 198 parcels out of 581 (table 8.9.) The graph of figure 8.23 indicates that after approximately 15 seconds, XUDP resorted to only sending the RELIABLE parcels. XUDP fell 7 seconds behind due to the required reliable data transfer of 1 parcel a second. With a properly programmed multimedia sending application, XUDP would perform quite a bit better.
|Total Data Acknowledged||3009.960 KBytes|
|Total Transmission Time||39.299 seconds|
|Network Bandwidth Utilization||76.591 KBytes/second|
|Average Round Trip Time||46.1 milliseconds|
|Average Window Size||9.8 packets|
|Total Parcels Received||198 parcels|
|Total Transmission Time||36.922 seconds|
|Total Parcels Skipped||383 parcels|
|Network Bandwidth Utilization (14.4 KBytes/parcel)||77.222 KBytes/second|
|Average Parcel Reception Frequency||5.363 parcels/second|
TCP's transmission of the 30 second video stream completed in 43 seconds, with an average bandwidth utilization of 195KBytes/second (table 8.10), approximately 30KBytes/second lower than that of the unloaded network test.
|Total Parcels Received||581 parcels|
|Total Transmission Time||42.986 seconds|
|Total Parcels Skipped||0 parcels|
|Network Bandwidth Utilization (14.4 KBytes/parcel)||194.631 KBytes/second|
|Average Parcel Reception Frequency||13.516 parcels/second|
The comparison graph in figure 8.28 shows that XUDP transmitted a good deal less data than TCP, and did a better job than TCP at preserving the linear progression of the data in time, but still not as good a job as would be desirable.