An information centric network for computing the distribution of computations

Abstract

Named Function Networking (NFN) extends classic Information Centric Networking (ICN), such that in addition to resolving data access by name, it also supports the concept of function definition and application to data (or other functions) in the same resolution-by-name process. This empowers the network to select internally (optimal) places for fulfilling a potentially complex user expression. Forwarding optimization and routing policies become thereafter a basis of dynamic decisions for (re)-distributing computations, and retrieving results. In this paper we describe the intrinsic operations and mechanisms of an instantiation of NFN based on untyped Lambda expressions and Scala procedures. Then, we demonstrate through a series of proof-of-concept experiments how they extend the capabilities of an information centric network (CCN), for orchestrating and distributing data computations, and re-using cached results from previous computations. In the end we report and discuss the main observations stemming from these experiments and highlight important insights that can impact the architecting of ICN protocols that focus on named-data.