Incorporating MPLS for Better SoC Utilization and Traffic Engineering

Abstract

Multiprotocol Label Switching (MPLS) traffic engineering by using SoC is discussed in this study. This study firstly provides a brief review of MPLS, constrained-based routing and OSPF protocol in order to create a background for traffic engineering. MPLS is used to provide a smooth flow of traffic into the network. In MPLS, Label Switch Router (LSR) uses a signalling protocol such as Label Distribution Protocol (LDP), Border Gateway Protocol (BGP) and Resource Reservation Protocol (RSVP) to establish label switched paths (LSPs). LSPs only follow explicit paths rather than following the shortest paths. In addition to allowing high-performance forwarding architecture, MPLS technology enables other important applications such as the ability to direct data traffic on the system on the chip around the network. However, MPLS is not a widely used technique in the SoC areas. MPLS is popular in the network engineering field but not much research is done yet on the semiconductor ICs. Fired by shared NOC bus is a single point of failure as well as has scalability issues, and on the other hand, high costs of fabricating network on chip (NoC) using a huge number of multiplexers (MUX components), our main goal is to implement and benchmark the suitability of MPLS traffic engineering on an integrated chip level. In this work, three different topologies and their simulation results are presented. The weaknesses of each topology and potential benefits are also tackled.