Traffic Reduction Technologies and Data Aggregation Control to Minimize Latency in IoT Systems

Hideaki YOSHINO  Kenko OTA  Takefumi HIRAGURI  

Publication
IEICE TRANSACTIONS on Communications   Vol.E104-B   No.7   pp.706-715
Publication Date: 2021/07/01
Publicized: 2021/02/04
Online ISSN: 1745-1345
DOI: 10.1587/transcom.2020CQI0002
Type of Manuscript: INVITED PAPER (Special Section on Future Directions of Research and Development on Communication Quality)
Category: 
Keyword: 
IoT,  fog,  edge,  data aggregation,  QoS,  latency,  control,  communication quality,  communication traffic,  

Full Text: FreePDF


Summary: 
The spread of the Internet of Things (IoT) has led to the generation of large amounts of data, requiring massive communication, computing, and storage resources. Cloud computing plays an important role in realizing most IoT applications classified as massive machine type communication and cyber-physical control applications in vertical domains. To handle the increasing amount of IoT data, it is important to reduce the traffic concentrated in the cloud by distributing the computing and storage resources to the network edge side and to suppress the latency of the IoT applications. In this paper, we first present a recent literature review on fog/edge computing and data aggregation as representative traffic reduction technologies for efficiently utilizing communication, computing, and storage resources in IoT systems, and then focus on data aggregation control minimizing the latency in an IoT gateway. We then present a unified modeling for statistical and nonstatistical data aggregation and analyze its latency. We analytically derive the Laplace-Stieltjes transform and average of the stationary distribution of the latency and approximate the average latency; we subsequently apply it to an adaptive aggregation number control for the time-variant data arrival. The transient traffic characteristics, that is, the absorption of traffic fluctuations realizing a stable optimal latency, were clarified through a simulation with a time-variant Poisson input and non-Poisson inputs, such as a Beta input, which is a typical IoT traffic model.