Aerospace and Electronic Systems Magazine April 2017 - 54
Integration of Reed-Solomon Codes to Licklider Transmission Protocol (LTP) for Space DTN
size and channel quality vary. This is because the corrupted segments are restored at the receiver by the RS error correction capability, following the decoding procedures described earlier. As a result,
the entire bundle is reconstructed at the receiver without requests for
retransmission of the corrupted segments. This leads to consistently
short bundle delivery time for RS-LTP as observed in Figure 3.
In summary, for the experimented scenarios, high data loss over
a space channel causes LTP-UDP to require additional transmissions rounds due to its sole reliance on ARQ-based retransmission.
This leads to its significant increase in data delivery time. In contrast, RS-LTP always successfully complete the bundle delivery
with only a single transmission round regardless of the rate of data
loss and bundle size, because of RS error correction. As a result, the
data delivery time is contributed only by the round-trip time and the
processing time involved with RS encoding and decoding to LTP
segments, leading to its short and nearly consistent delivery time.
In some space communication scenarios, goodput performance, defined as a ratio of the unique number of delivered application data bytes to the total data delivery time, is needed as a
measure of the transmission efficiency and link utilization for a
protocol. To evaluate the effect of bundle size and channel quality
on goodput performance of RS-LTP, we provide in Figure 6, a sample comparison of goodput performance from experiments for the
transmission with the bundles of 8 Kbytes and 32 Kbytes. As observed, corresponding to the comparisons of the data delivery time
between the bundles of 8 Kbytes and 32 Kbytes in Figures 3(b) and
3(d), RS-LTP outperforms LTP-UDP regardless of the bundle size
and channel BER. The goodput advantages of RS-LTP over LTPUDP increase along with the increase of BER. This is because the
goodput of LTP-UDP drops severely for the increase of BER while
it remains nearly unchanged for RS-LTP at all three BERs.
The goopdut performance variations of both protocols in Figure 6 are reasonable. As mentioned, for all the transmissions of
LTP-UDP with individual bundle size, the total data delivery time
increases as the channel BER increases. As goodput is defined as a
ratio of delivered unique data bytes to the total data delivery time,
there must be a corresponding decline in the goodput for LTP-UDP
along with the increase of BER, reflecting its increase of data delivery time in Figure 3. Similarly, for the transmission of RS-LTP,
because of the nearly consistent data delivery time regardless of
the channel BER and bundle size, it has nearly equal goodput performance all the time as observed from Figure 6.
CONCLUSIONS AND DISCUSSIONS
In this article, RS-LTP is proposed as an implementation of RS
erasure codes in a new "local data-link layer" protocol to support
DTN's primary data transport protocol, LTP, for more efficient data
delivery in space networks. The performance evaluation based on
data transfer experiments over a PC-based testbed reveal that, regardless of DTN bundle size, RS-LTP shows significant performance advantages (for both data delivery time and goodput) over
LTP-UDP (LTP with simple, non-encoded UDP at the local datalink layer) for transmission over a cislunar space channel with the
channel BERs ranging from 10−6 to 10−4. The performance variation trend is that the larger the bundle size and/or the higher the
Comparison of the goodput performance between LTP-UDP and
RS-LTP for transmission with different channel qualities for different
bundle sizes. (a) Bundle size = 8 Kbytes. (b) Bundle size = 32 Kbytes.
channel error rate, the greater the performance advantage of RSLTP over LTP-UDP. It is also found that variation in channel quality (i.e., channel error rate) has significant effect on the transmission performance of LTP-UDP while it has almost no effect on
RS-LTP. This implies that RS-LTP is much more tolerant of change
in channel transmission conditions than LTP-UDP.
The significant performance advantages of RS-LTP over LTPUDP are explained by the different mechanisms implemented for
two protocols to recover for data loss. For LTP-UDP which has no
RS coding implemented, its loss recovery solely relies on LTP's
ARQ-based retransmission mechanism. Therefore, in the event of
data corruption over a lossy channel, the corrupted data segments
are simply retransmitted, which results in additional transmission
rounds for successful delivery of an entire bundle. This leads to
much longer bundle delivery time and thus severe performance
degradation for the LTP-UDP transmission over a lossy channel.
In contrast, for RS-LTP which implements RS error correction, the
IEEE A&E SYSTEMS MAGAZINE