Aerospace and Electronic Systems Magazine April 2017 - 53
Shi et al.
Comparison of the measured number of transmission rounds between
LTP-UDP and RS-LTP for successful delivery of data bundles at different bundle sizes and channel quality settings.
Comparison of data delivery time among three channel BERs at four
experimented bundle sizes. (a) Comparison for LTP-UDP. (b) Comparison for RS-LTP.
observed that the data delivery time with the BER of 10−4 increases more significantly along with an increase in bundle size than
other two transmissions. Numerically, the increase of the bundle
size from 4 Kbytes to 32 Kbytes leads to an increase of the data
delivery time for around 10 sec while it is only around 5 sec and 2
sec for the BERs of 10−5 and of 10−6, respectively. In contrast, the
data delivery time of RS-LTP remains nearly constant regardless of
the channel quality and bundle size, with only very minor increase
observed at the BER of 10−4 for the bundle size of 32 Kbytes.
The performance differences with respect to the change in BER
between LTP-UDP and RS-LTP observed in Figure 4 imply that,
depending on the choice of "local data-link layer" protocol (LTPUDP or RS-LTP), variation in channel quality will have different
effects on data transmission performance-channel quality variation has significant effect on the transmission performance of LTPUDP but it has almost no impact on the performance of RS-LTP.
To investigate the reasons that caused the performance differences between two protocols, we measured the number of transmission rounds experienced for successful delivery of an entire
bundle in each case. Figure 5 presents a comparison of the measured number of transmission rounds for successful data delivery
between LTP-UDP and RS-LTP for all four bundle sizes and different channel quality settings. Note first that, at the lowest channel
error rate (i.e., a BER of 10−6), LTP-UDP needs two transmission
rounds to deliver all the segments of a bundle to the receiver while
RS-LTP needs only one round to complete the entire data delivery.
As the BER increases, the number of transmission rounds required for successful data delivery of LTP-UDP increase for all
four bundle sizes: it increases slightly (from 2 to 3 or 4) when the
channel BER increases from 10−6 to 10−5, but it increases sharply
to 6 or 8 (depending on the bundle size) when the BER further
increases to 10−4. For a given bundle size, the higher the channel
BER, the more corruption events are likely to occur. For LTP-UDP
which relies solely on ARQ-based retransmission strategy for reliable delivery, the sender simply retransmits the data segments corrupted or lost during transmission in response to a request from the
receiver. Even the corruption of a single bit must result in retransmission of the entire segment. Therefore, more corruption events
lead to an increase in the number of retransmission rounds required
for successful delivery of an entire bundle. Given that an additional
RTT of 2.7 sec is involved for each transmission round, this leads
to much longer data bundle delivery time for LTP-UDP transmission. This is why the data delivery time of LTP-UDP increases with
an increase of channel BER for a constant bundle size in Fiure.
3. A large bundle surely aggravates the situation leading to more
retransmission rounds for LTP-UDP and thus, longer data delivery
time as observed for the bundle size of 32 Kbytes at BERs of 10−5
and 10−4 in Fig. 4(a).
For RS-LTP, the same number of segments are lost or corrupted
during transmission as for LTP-UDP because the transmission conditions are the same. However, as shown in Figure 5, the number of
required transmission rounds remains 1 no matter how the bundle
IEEE A&E SYSTEMS MAGAZINE