By analyzing the keywords of edge computation-related papers published in mainstream academic institutions such as IEEE/ACM/USENIX, we can get the statistical results shown in Figure 1-9. It can be seen that in the research work of edge computing, the Internet of Things, computing offloading, resource allocation, 5G, and deep learning are the top five keywords. This also basically covers the research trends and recent advances in edge computing.
NE870 Figure 1-9 Relative word frequency of keywords in edge computing-related articles in the last 5 years
01 Internet of Things + Edge Computing
The Internet of Things has become the first hot word in edge computing related research papers, reflecting the close relationship between edge computing and the Internet of Things. The more mature the development of Internet of Things technology, the stronger the technical demand for edge computing. The key to the combination of the two lies in two aspects:
On the one hand, there is a need to address how iot devices can access edge computing in a low-cost way;
On the other hand, it is also necessary to solve how edge computing can cope with the characteristics of massive, heterogeneous and dynamic Internet of Things services.
02 5G+ Edge computing
The development of 5G technology will enable communication latency to reach a level lower than computing latency, which will make many existing computing models fundamentally change, but also lead to more and more computing load from front-end mobile devices to edge computing servers.
This puts forward new requirements and challenges for the architecture of edge computing, and it is necessary to make important improvements on the basis of the existing cloud computing cluster architecture to meet the needs of 5G mobile services with high real-time, data-intensive, high mobility, and heterogeneous dynamics.
03 Virtualization Technology
Due to the heterogeneity of front-end devices, the computing requests served NE870 by edge computing are also highly heterogeneous. This requires the edge server to be able to run a variety of computing services flexibly. Virtualization technology is one of the mainstream directions to solve this problem, through different systems, different environments and even different hardware network functions on the common computing resources to achieve flexible management of network functions.
Compared with the virtualization technology in traditional cloud computing, the virtualization technology of edge computing has higher latency requirements. Not only that, edge servers also have much less computing resources than cloud servers, making virtualization technology as lightweight as possible.
04 Computing Uninstallation
Computing offload is one of the classic problems in cloud computing, and it is also a very important core problem in edge computing.
Computing offloading in edge computing refers to the transfer of computing tasks from the front-end device to the edge server, and then the edge server returns the computing results to the front-end device or delivers them to the cloud server as required. Research in this direction has focused on answering a few key core questions – whether to uninstall, which tasks to uninstall, to which server, and how to uninstall.
Compared with task unloading in cloud computing, an important feature of edge computing is the transmission mode of front-end devices and the selection of edge servers, which will seriously affect the performance of computing unloading.
05 Resource Allocation
There may be a large number NE870 of edge servers in the same edge computing network, and the same edge server may need to process a large number of computing tasks, and different computing tasks have different computing and communication resource requirements. Based on this, resource allocation in edge computing becomes particularly important.
Unlike cloud computing data centers, edge computing is closer to front-end users, and the services it runs and the resources it is equipped with are highly targeted. Moreover, resources on different edge servers are often highly heterogeneous, which makes resource allocation in edge computing extremely challenging.
06 Low power iot systems that support edge computing
The proposed edge computing is not aimed at specific application scenarios, but plays a more similar role to the content delivery network to reduce the access delay of applications. This feature can solve the problems of limited energy and limited resources of the Internet of Things system. NE870 In addition to the exploration of various applications, the common problems in this direction include low-power embedded systems (supporting compute offloading, low-power task transfer, energy-efficient data acquisition, etc.).
07 Edge computing and artificial intelligence algorithms
The collision of edge computing and artificial intelligence produces a series of problems in two directions, namely, the artificial intelligence algorithm based on edge computing and the edge system optimization based on artificial intelligence. Compared with the traditional artificial intelligence algorithm, the change of the former system architecture brings the problem of collaboration between multiple devices. The latter is to use the data generated by artificial intelligence algorithms and edge computing system processes to optimize and make decisions on the edge system itself.
Considering that one of the important missions of edge computing is to bring artificial intelligence into various iot devices, this direction is attracting increasing attention.