drastic. Algorithm (c), the tasks are not prioritized and all are executed on the MEC
server, resulting in high-priority tasks waiting for a long time. Our algorithm is
dynamically selected through the current network situation, which has the best performance effect.
Notice the intersection of algorithm (a) and algorithms (b) and (c), and when the
number of devices reaches a certain upper limit, all tasks take longer to execute
remotely than tasks execute locally. From the formula, we can easily analyze the
reasons.
T
l
k ¼ t
tr
k;s )
CW k
f l
k
À
CW k
f
c
k
¼ CW k Ã
f
c
k À f
l
k
f l
k f
c
k
¼
DS k
R k;s
þ t
w
k
ð3:1Þ
3.2 Average Execution Time for Different Priorities
Secondly, we investigate the changes in the average execution time of tasks under
different priorities, as shown in Fig. 5. Algorithm (a), the processing time of priority
tasks is almost the same and must be greater than our algorithm. Algorithm (b), for
those with high priority the time consumed is less than executed locally, and it takes
Table 2. Second filtering algorithm description
Research on Multi-priority Task Scheduling Algorithms …
145