Dedicated Channels as an Optimal Network Support for Effective Transfer of Massive Data

Abstract

Instantaneous fair sharing (IFS) is a traditional network ideal prescribing to share the network capacity among competing applications fairly during any infinitesimal time interval. In this paper, we argue that IFS is an inappropriate ideal for the application of massive data transfers where the primary goal is to minimize message transfer times. We propose an alternative paradigm of virtual finish time first (ViFi) scheduling that dedicates the entire capacity to one message at a time in the order of message finish times under IFS. Unlike shortest remaining time first and other earlier algorithms for dedicated scheduling, ViFi provides a remarkable guarantee of delivering each message no later than under IFS. Our analysis and simulations show the dedicated ViFi scheduling offers significant reductions in the average transfer time. The above properties make ViFi a promising approach for resource allocation in emerging dedicated-channel networks that enable advance reservation of end-to-end channels between hosts.