IEEE Consumer Electronics Magazine - January/February 2023 - 55

TABLE 2. Simulation parameters.
Parameter
Simulation area
Stationary node
Broker version
Dew Buffer Size
Message TTL
QoS level
Mobile node
Value
3000 m2
30
1.6.7
50-250 MB
300 min
1,2
The subscriber of the message chooses minimum
timestamp values amongst all encountered
messages, which we can present as
tsub ¼ minð tm nðÞ1;tm nðÞ2 ... ;tm nðÞnÞ:
(4)
B. End-to-End Latency Formulation
Here ld is the latency for transmitting the data
from the sensor node to the dew-node via the
gateway and dew-broker by considering MQTT
QoS 1 level. Also, we can formulate the latency
for the transmission from the dew-node to the
edge-node as le. Therefore, the total latency, in
this case, is computed as follows:
ldb ¼ ltx þ lcnack þ lpuback þ lfwd
:
(5)
Here;ldb is the latency from the sensor to the
broker. In addition, we can formulate ldn as
latency from the broker to the dew-node
ldn ¼ lcnack þ lpubrel þ lsuback þ lfwd
:
sor to cache node can be expressed as
ld ¼ ldb þ ldn:
(6)
Therefore total latency achieved during sen(7)
We
can also formulate the latency for data
transmission towards edge-node le through the
edge broker by repeating the same process.
Therefore, we can write
le ¼ leb þ len
(8)
where leb ldb and len ldn.
Finally, we can derive the total latency L for
the IoDT node as the sum of the latency of all
Dew and the edge node in the network as
L ¼ ½ld iðÞ þ le iðÞ :
i ¼ 1
Xn
(9)
RESULT ANALYSIS
We simulate the data collection from the
stationary node in two-phase using 802.11
January/February 2023
interfaces with Opportunistic MQTT protocol.
We have used the opportunistic network environment
simulator to validate the opportunistic
message transfer. We validate the CPU and memory
utilization in a standalone Raspberry Pi
testbed. The main advantage of MQTT compared
to Google RPC is the varying degree of QoS levels
and the low data footprint. We can use network
slicing to make the different clusters for edge
and dew networks in such a scenario. The message
transfer, in this case, has been performed
using QoS 2 MQTT service.
We have set the edge node as subscribers,
and the dew node, in this case, performs caching
of the message. Figure 3 shows the publisher-
subscriber message and acknowledgment transfer
sequences for the sensor node to the edge
nodes. The experimental simulation parameters
we tabulated in Table 2.
Message Delivery Performance With Varying
Buffer Size
Figure 4 represents the message delivery performance
(DP) of the various message transfer
scenarios for the dew network. A well-known
message transfer scheme has been addressed
and compared in this scenario, namely the
opportunistic flooding, probabilistic forwarding,
and opportunistic MQTT, the re-engineered
MQTT mechanism.13 From the figure, it is clear
that the opportunistic MQTT outperforms the
rest, and it goes to the highest benchmark of DP
while the buffer size is 150 MB. Because the message
transfer mechanism leverages the tokenbased
forwarding approach, which ensures high
confidence in message transfer.
A larger buffer size, in this case, results in a
low packet drop in the network since the
55
Figure 4. Buffer size versus DP performance of
message transfer mechanisms for the proposed
ecosystem.
70
IEEE Consumer Electronics Magazine - January/February 2023

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - January/February 2023
