Flow assignment method with traffic characteristics over multiple paths for reducing queuing delay

Abstract

In traffic engineering (TE), it is vital to take traffic characteristics of the flows into account in appropriately assigning the flows to multiple network paths to achieve better delay performance as a whole in order to effectively distribute traffic flows over the paths. This paper presents a novel traffic characteristic-aware flow assignment method to reduce the queuing delay in a fundamental case where two types of flows with distinct traffic characteristics (e.g., burstiness) are distributed into two paths. First, we extensively analyze the queuing delays in assigning flows in the manner of various combinations of flows in terms of minimizing the worst queuing delay among two paths and show that it is not easy to find the optimal flow assignment when the paths have different bandwidths. Second, we propose an on-line flow assignment method for the different-Y. Kitatsuji ( ) bandwidth paths and show that the numerical simulation with the method finds a nearly optimal flow assignment and outperforms up to 40% compared with the conventional path-bandwidth-based flow assignment. Our evaluation suggests that considering the traffic characteristics in the flow distribution over multiple paths significantly improves the delay performance when the flows have distinct characteristics.

A Design and Prototyping of In-Network Processing Platform to Enable Adaptive Network Services

“…In our research group, a flow-based routing scheme [36] and a packet early discard scheme [37] have been considered for delay-sensitive application flows. In general, temporal and spatial resource optimization has been actively studied, e.g., rate and retransmission controls on multi-path routing, network coding, compression, and/or data caching.…”

Advanced Relay Nodes for Adaptive Network Services - Concept and Prototype Experiment

“…In our research group, flow-based routing scheme [15] and packet early discard scheme [16] have been considered for delaysensitive application flows. In general, temporal and spatial resource optimization has been actively studied: e.g., rate and retransmission controls on multi-path routing, network coding, compression, and/or data cache.…”