Simulation Results
Evaluating the proposed emergency management system involves many influencing factors, and complicated relationships exist among them. In the simulation experiments, we focus on each module’s response delay and the accuracy of the prediction system workload, which are the key factors affecting system performance.Our cloud-based evacuation system includes five machines to process the smartphone and sensor data. We set up a levee collapse context with 150
request senders as smartphone devices with random locations onsite. We treat the multimedia analysis work as MapReduce tasks, assigning one machine, the JobTracker, to read the data from the smartphone and allocate it to the other four machines, or TaskTrackers. At the same time, the 40 strength sensors and 40 water-level sensors placed on the levee report the strength value to the datacenter every 10 seconds. We assign the workload equally to two
machines in parallel.