Novel orchestration architecture for Fog computing

Cloud computing technologies such as virtualization and orchestration have revolutionized a great number of different domains, from the industry to healthcare. However, there are still certain applications that can not take advantage of this paradigm due to stricter requirements such as end-to-end response times. A new paradigm called Fog computing is arising to answer the needs of these applications. Fog nodes have lower hardware resources that their Cloud counterparts and are distributed over a wider geographical area. Porting technologies designed for the Cloud, specifically orchestration, to this new paradigm requires to handle these peculiarities. A novel orchestration architecture is proposed evolving on current ones, directly focusing Fog characteristics. Multiple schedulers are dynamically deployed in order to achieve a better scheduling throughput and spread the resources required by the scheduler among multiple lower-resource nodes. It uses an application-centric approach at the scheduler level in order to enable new scheduling parameters. The scheduling throughput of this architecture is compared to the most common Cloud orchestration architecture.