Edge task allocation scheme based on data classification

As the fifth-generation mobile communication technology matures, more and more task data is generated at the edge of the Internet. Edge computing has received extensive attention in the commercial and scientific fields due to its low cost, high efficiency and data privacy. However, considering that the types of tasks generated by the Internet of Things are more diversified as the user needs grow, the arrival time of different types of tasks is irregular, and the task allocation scheme is not flexible when the computing resources in the edge are relatively fixed. This paper proposes a method for dynamic task allocation using data classification in edge environment. The core framework of the method consists of three parts: task classification model, task quantity prediction, and task allocation strategy. In task classification model, using the similarity of feature attributes between tasks, the approach can effectively classify the arrived tasks into specified sorts. For each sort, the time series prediction method predicts the amounts of tasks arriving at the next moment, providing a reference for subsequent task assignment. Then, an efficient task allocation strategy was designed. Dynamically provide computationally-matched computing resources for tasks in each sort to meet their mission requirements. Finally, simulation experiments were conducted in the Google Cloud tracking data set. The results show that the method maintains a high completion rate compared with the traditional method when the number of tasks varies greatly. This proves the validity and robustness of the proposed dynamic task allocation strategy.