Abstract

Electrical-to-optical domain conversions and vice versa (denoted by O/E/O conversions) for each hop in optical core transport networks impose considerable capital and financial overhead on the providers. In this paper, we propose a full-mesh topology driven core network with a central scheduler that handles the task of signaling and coordination among slot transmissions for every hop to eliminate such O/E/O conversions. We introduce the concept of a container as a macro data unit that forms a separate layer in the protocol stack above the optical layer. A FAST centralized scheduling algorithm is proposed based on a preemptive scheduling technique that can ensure that there are no collisions between the containers. We also analyze the complexity of this algorithm. Next we design the logical architecture for the core and edge switches following the de facto policy of moving the complexity to the edge. We also designed a hierarchical architecture for the edge switch and provide the respective block diagrams. To get a more concrete design prototype, we further proposed a generic (vendor independent) physical architecture for a single port of the switch considering SONET/SDH on the access side. Moreover, we develop a concise delay model for the containers to analyze the packet arrival process and derive the optimal container size, based on the link speed. Finally, we present some simulation results to study the performance of the algorithms and models proposed in our work.

Keywords: Photonic network; Edge router architecture; Central scheduler; G/D/1 queue; Stochastic delay model

Article Outline

1. Introduction

2. Architecture of the photonic network and component tasks

2.1. Control and data planes

2.2. Edge and core nodes

2.3. Components of the architecture

2.4. High level view of containerization in the switch

2.5. Hierarchical architecture

2.6. Interdisciplinary network processors

2.7. Data flow through the access and network sides

2.8. Mapping between functional and physical architecture of the switch

3. Centralized scheduler design

3.1. Static scheduler with traffic forecasting

3.2. The scheduling problem

3.3. Schedule optimization and lower bounds

4. Container and delay model

4.1. Analysis of container formation

4.1.1. Quantization

4.1.2. Arrival traffic assumption

4.1.3. Container formation tree

4.1.4. Tree collapse

4.1.5. Container tree algorithm

4.1.6. Container queue delays

4.2. Container queue delay approximation

5. Numerical results

6. Conclusion

References

Vitae