Abstract
Methodologies for Synthesizing and Analyzing Dynamic Dataflow Programs in Heterogeneous Systems for Edge Computing
Highlights
T HE SUCCESS of the Internet of Things (IoT), and the growing trend of implementing decentralized processing applications close to the data generation devices, has created the needs for efficient edge computing approaches, in which data processing occurs partially already at the network edge, rather than in the cloud [1]–[4]
WORK A new methodology based on dataflow programming for the porting on heterogeneous CPU/GPU processing nodes is described in this paper
The results have been achieved by extending the formalism already developed by the authors, based on the modelling, analysis and automatic generation of application source code from a high-level representation based on a dynamic dataflow computation model to include porting and partitioning on CPU/GPU nodes
Summary
T HE SUCCESS of the Internet of Things (IoT), and the growing trend of implementing decentralized processing applications close to the data generation devices, has created the needs for efficient edge computing approaches, in which data processing occurs partially (or completely) already at the network edge, rather than (completely) in the cloud [1]–[4]. Edge computing can be considered as a new computing paradigm in which the data is processed at the edge of the network, where the usage of cloud resources can be reduced (or even completely eliminated) [1] The problems that such new paradigm intends to solve covers several aspects: data access latency, limited battery life of mobile devices, limited amount and cost of bandwidth, security and privacy of interchanged or stored data are just some of the examples [2], [5], [6].
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
