Article Preview
Top1. Introduction
The unprecedent development of the Internet of Things (IoT) and the popularization of 5G network, has significantly increased the number of network edge devices. The centralized big data processing technology with cloud computing model as the core is not able to efficiently process the data generated by edge devices. Therefore, the edge data processing with the edge computing model as the core emerges as the need of the hour. The former generates a small amount of data computing, which is consistent with the existing centralized data processing with the latter as the core, and better solves the problems existing in the era of IoT.
In the cloud computing model, users write applications and deploy them to the cloud. User programs are usually written and compiled on the target platform, run on the cloud server, save or transfer data to the cloud, and finally process in the cloud. Based on this centralized data processing model, there are batch processing (Dean et al., 2008; Isard et al., 2007; Saha al., 2015; Zaharia al., 2010; Shvachko et al., 2010) and flow processing (Qian et al.,2013; Akidau et al., 2013; Apache storm 2016; Neumeyer et al., 2010; Kulkarni et al., 2015; IBM 2016; Zaharia et al., 2013) methods, and the application program can be centralized processing on the cloud platform. However, in the edge computing model, some or all the computing tasks are migrated from the cloud to the edge nodes, but most of the edge nodes are heterogeneous platforms, and the running environment of each node may be different. Therefore, while deploying user applications under the edge computing model, programmers encounter greater difficulties. The traditional programming method MapReduce, Spark and others are not suitable, so it is necessary to study a new programming method based on edge computing (Weisong et al., 2017).
In order to realize the programmability of edge computing, Hongjie et al. (2019) proposed a state perception model of building equipment based on edge computing. Xiong et al. (2019) studied the safety risk assessment method based on edge computing in the national grid. In terms of the task migration, Jia et al. (2019) suggested a task migration model for mobile edge computing. Xin-feng et al. (2019) advocated a dynamic resource allocation strategy in mobile edge computing environment. These models lack the programmable strategy of edge computing. Edge Computing: Platforms, Applications and Challenges (Ziming et al. 2018), and Near end Cloud Computing: opportunities and challenges in the post cloud computing era (Yuezhi et al. 2019), took programmability as the challenge of edge computing and showed the importance of edge computing programmability.
Hong et al. (2013) proposed a high-level programming model for future Internet applications, which has the characteristics of on-demand expansion, large-scale geographical location distribution and delay sensitivity. The programming model cloud aware proposed by Orsimi et al. (2015) estimates the connection state of the network through a context manager and splits the application and performs the uninstall by solving the optimization problem. Quan et al. (2016) suggested a new programming model based on edge computing firework model. Firework model is a kind of object-oriented programming model and has good generality. It extends the visualization boundary of data and provides a new programming model for data processing in collaborative edge environment (CEE). However, most of the current achievements are based on theoretical research, and the realization in the real environment is still unknown.