IEEE - Aerospace and Electronic Systems - May 2020 - 7

Shi et al.
and thus a monotonic increase of the RTT, the setting of CP
timer length is critical for transmission efficiency of LTP
over the deep-space communication channel.

EXPERIMENTAL SETUP AND CONFIGURATIONS
The analysis of the CP timer length presented in this article is
done based on the realistic file transfer experiments conducted using a PC-based testbed. The testbed infrastructure
adopted for the proposed file transfer is the PC-based space
communication and networking testbed [25]. The protocol
implementation, LTP together with BP, adopted for the
experiments was from the Interplanetary Overlay Network
(ION) distribution v3.6.2. Developed by NASA's JPL,
California Institute of Technology, the ION is a software
implementation of the DTN protocol suite intended for reliable and highly efficient data delivery in space networks and
deep-space communications [30].
A one-way delay of 1.35 s, which is common over a
cislunar channel, was introduced to both the data and
ACK channels to emulate the signal propagation delay in
a deep-space communication scenario. For the sake of
comparison, the experiments with both the symmetric
channel and asymmetric channel are studied. In the case
of symmetric channel, both a downlink channel rate and
an uplink channel rate are configured to be 2 Mbps.
In the case of asymmetric channel, the downlink channel rate is kept as 2 Mbps, whereas the uplink channel rate
is reduced to 4 Kbps. This leads to a high channel ratio
(CR) of 500/1 for the experiments. In other words, for the
transmission with asymmetric channels, the transmission
rate of LTP data blocks is 2 Mbps and the transmission
rate of RS segments in the opposite direction is 4 Kbps.
A text file of 1 Mbyte is transmitted by running LTP
between the sender and receiver over the testbed to measure
the performance of the protocol. With the bundle aggregation capability available with the ION distribution, the file
transmission is configured to have five bundles aggregated
within each LTP block. The length of each bundle is 1000
bytes. Given that the Ethernet is adopted to provide link layer
framing service in our experiments, the aggregated block is
fragmented as data segments with each segment of 1400
bytes. By this, the segment even after encapsulation process
at the LTP and IP layer, with a length of 1460 bytes, is still
able to fit into a frame MTU of 1500 bytes at the data link.
To ensure transmission reliability of each block, all the data
bytes of each segment are set as 100% red.

PERFORMANCE EVALUATION RESULTS AND
DISCUSSION
This section is dedicated to discussion of performance
evaluation results of the file transfer experiments. The
MAY 2020

discussion focuses on the effect analysis of the CP timer
lengths using the time sequence graphs (TSGs) and goodput graphs [31]. The impacts of both symmetric channel
and asymmetric channel configurations are also considered. A comparison of the goodput performance among
multiple CP timers is also presented.

ANALYSIS OF CP TIMER LENGTHS FOR LTP
TRANSMISSIONS OVER SYMMETRIC CHANNEL
In Figure 1, a TSG is presented to illustrate the LTP transmission at the packet (or segment) level for delivery of a
file of 1 Mbyte over an emulated symmetric cislunar channel with a BER of 0 experienced. The TSG is generated
using traffic dumped at the sending node for file delivery.
Provided that the one-way link delay of 1.35 s is configured, the CP timer length is set to 4 s, which is greater
than the estimated RTT.
The TSG variation trend shows that the entire file of
1 Mbyte is transmitted in a linearly increasing time
sequence of data segments with continuous LTP block
transmission. Acknowledging the successful receipt of all
the blocks to the sender, the corresponding RSs from the
data receiver are also transmitted in the same linearlyincreased pace as the blocks. This leads to nearly consistent length of RTT for each of the blocks.
The operation of the LTP's bundle aggregation and
one RS (i.e., ACK) per block mechanism can be viewed
from the TSG at packet level. Figure 2 shows an enlarged
view of the TSG for transmission of two blocks in the
middle of the file. As mentioned, the file transmission is
operated with five bundles aggregated within each LTP
block, with the configured bundle size of 1000 bytes and
the size of each aggregated block of 5000 bytes. For the
block fragmentation, with each encapsulated segment of
1460 bytes, each block should be divided into four segments according to d5000
1460e. The last one out of all four segments is expected to carry slightly fewer number of data
bytes than other three. This is exactly what the data transmission line of the TSG shows in Figure 2. Following the
one RS per block policy, the entire block of four segments
is acknowledged by a single RS segment as illustrated.
The time gap (over the x-axis) between each block and
its corresponding RS segment in Figure 2 is around 2.8 s.
This is consistently the measured RTT interval for these two
blocks and all other blocks of the file. This is reasonably
given that the one-way propagation link delay is set as 1.35 s
with other time components, such as processing time and
possible queue time considered. Because the CP timer length
(of 4 s) is longer than the estimated RTT of 2.8 s for each
block, the RS is received by the block sender, and thus no
need to retransmit the CP segment. Therefore, there is no
retransmission of any segment observed between the TSG
data line and RS line in Figures 1 and 2.

IEEE A&E SYSTEMS MAGAZINE

IEEE - Aerospace and Electronic Systems - May 2020

Table of Contents for the Digital Edition of IEEE - Aerospace and Electronic Systems - May 2020