Data-Centric Edge Federation: A Multi-Edge Architecture for Data Stream Processing of IoT Applications

Emerging Internet of Things (IoT) applications demand data stream processing with low latency and high processing power. Although the cloud naturally provides huge processing capacity, high latency to move data to the datacenter is prohibitive. Edge computing is a recent paradigm where part of computing and storage resources are pushed from the cloud to the edge of the network. In edge computing, edge providers manage their resources near to IoT devices to meet low latency application requirements and reduce the network core bandwidth. To reach the maximum potential of edge computing, a big challenge is to promote the cooperation between edge providers. Currently, edge computing architectures are severely limited for providing cooperation mechanisms between distinct edge providers. In this paper, we propose a edge federation to leverage the cooperation between different edge providers. The edge federation uses interest information propagated in data streams that travel between edge providers to allow an stakeholder to react to inefficient resource allocation and service provision. The main objective of the federation is to create a consortium of edge providers to provide cooperation mechanisms and define and standardize the application interests. The proposed edge federation is (i) data-centric, since edge providers can share common interests and data and, thus, establish cooperation to increase the capacity to provide services for applications; (ii) distributed, since no assumption is made concerning the geo-location of the edge providers and their logical connections; (iii) opportunistic, because an edge provider can react dynamically to the environment change ; (iv) scalable, since the edge provider has the ability to analyze a data flow passing by its infrastructure and make decisions to increase network performance locally, which impacts the global performance