Page 80 - RAC_CIAW_ a_I_n_01_2021.pdf

running isolated in its container. asymmetric cryptography. Figure 4a shows CPU and
To simulate a Naval District blockchain server, we memory consumption during the process. The red line
virtualize a Linux server by the VirtualBox Hypervi- represents the CPU consumption by the SSH protocol,
sor. In the Linux server, we virtualized each blockchain reaching an average of 8% and standard deviation of
Peer in Docker’s containers. Each container has its own 5%. The ligth green line represent the CPU con-
network interface, enabling its direct communication sumption by the OS, with an average of consumption
to the others entities and to the sensing nodes. The pro- of 7% during the transmission process. The memory
totype code and all applications necessary to run the consumption stays stabilized around 156 MB.
experiment are available in a GitHub public reposito- Figure 4b shows the client performance while
3
ry . Our virtualized Linux server is set with an Intel sending the 1500 AIS entries via blockchain’s transac-
Core i5 (1.8 GHz) and 2048 MB of memory. The cli- tions. As expected, we can observe the increase of both
ent’s Raspberry Pi 3 has a Cortex-A53 Quad-core (1.4 CPU and memory consumption. The magenta line rep-
GHz) processor and 1048 MB of memory. resents the blockchain’s client CPU consumption, with
We positioned the AIS VHF antenna near Guana- an average of 17% and a standard deviation of 8%,
bara Bay, Rio de Janeiro, to receive the AIS data of however, without compromise the OS performance,
the area marine traffic. The AIS data broadcast by that remains approximately the same with an average
the ships are received by the VHF whip antena and of 6%. Memory consumption stabilizes around 180
analog-to-digital converted by the SDR dongle. The MB during the transmission.
SDR dongle control program then save each entries in Our comparison shows a blockchain overhead of
a text file that is read by the client application. To sim- 9% in CPU usage and 14.7% in memory usage over
ulate an MMS with a vast number of sensing nodes the SSH protocol. Nonetheless, the blockchain client
and a large amount of sensing data, we send 1500 AIS uses only 17.6% of all 1024MB of memory and 16.9%
entries, representing 100 nodes sending 15 AIS entries of CPU in the Raspberry Pi 3. It’s also important to
in a minute. highlight that the blockchain transmission was 60%
First, we compare the client performance eval- faster, taking about 60 seconds, while the SSH trans-
uating the CPU and memory usage by sending 1500 mission last around 150 seconds.
AIS entries through the Secure Shell (SSH) protocol The client end was one of the main concerns in
and then through blockchain. We chose SSH due to our research because these sensing devices are often
fact that the protocol employs both symmetric and installed in platforms as buoys and UAVs that can’t
support heavy payloads. Besides that, these devices
3 Available at: www.github.com/warfreire/blockchain-based.sen- need to be prepared for heavy weather conditions and
sing.system

(a) Sending data via SSH (b) Sending data via blockchain’s transactions

Fig. 4: Sensing node sending 1500 AIS entries

80 REVISTA ACADÊMICA CIENTÍFICA DO CIAW