Manivasakan Rathinam

With the emergence of IoT, the number of devices that is getting connected is increasing exponentially, which poses a constraint on the lower latency requirements of processing these tasks. Evolving technologies such as fog computing and edge computing consist of computationally lesser exhaustive power servers, which bring this processing near the edge or onto the devices, thereby reducing the round-trip delay as well as the load on the entire network. In this work, we propose the hierarchical arrangement of servers in the fog computing layer for this processing of data and derive the analytical framework for the same. The proposed architecture works on top-down scheduling policy rule for the incoming packets, defined mathematically in terms of two-dimensional Markov chains. The performance of the proposed architecture is compared with an equivalent flat architecture analytically in terms of the mean sojourn time and the mean computational power and justified using simulation results.

Next-generation wireless networks are envisaged to provide end-users ubiquitous low-latency computing services using devices at the network edge and machines before reaching the core cloud network. The crucial concepts of driving this technology are to offload computationally intensive tasks from users to edge or fog devices. The mobile edge network consists of an access point (AP), Radio Access Network (RAN), devices, edge servers, fog servers and finally the core cloud network. Given that fog computing is an emerging technology, it is imperative to study the performance of such systems analytically. Though there are many papers on performance analysis, many of them provide heuristic, ad-hoc solutions or pose it as optimization problems. In this work, we model a hierarchical fog architecture using Markov chain analysis and compare it with a flat fog architecture by investigating the different performance measures analytically and by simulation.