Yohei Iizawa

Multi-domain ASON/GMPLS network operation: current status and future evolution

This paper describes clarifications and comparisons of routing models for multi-domain optical networks as well as the current status and future perspective for ASON/GMPLS technologies. The three routing models, the per-domain routing, the ASON hierarchical routing, and the PCE-based routing, have been being standardized for the ASON/GMPLS multi-domain networks. In this paper, we compare these three routing models in terms of path computation capabilities, on-line multiple path planning capability and inter-domain confidentiality. According to our analyses, the PCE-based routing and the ASON hierarchical routing models have rich path computation capabilities enough for multi-domain ASON/GMPLS network operation, while the per-domain routing does not provide much capabilities. One of the significant issues left for the ASON hierarchical routing model is an appropriate abstraction mechanism which strongly affects the path computation capabilities and inter-domain confidentiality. In addition to the sufficient path computation capabilities, the PCE-based routing model offers the on-line bulk optimization for multiple requests. Therefore, we concluded that the PCE-based routing is the most suitable for multi-domain ASON/GMPLS networks.

PCE-based Scalable Dynamic Path Control for Large-scale Photonic Networks

This paper proposes an inter-domain path control system based on PCE (Path Computation Element) for large-scale photonic networks, especially for a photonic network across over multiple carrier domains. We discuss the overall architecture of multi-domain optical network control system. The system utilizes maximum flow information in path computation and Path Key scheme in signaling, which improve the load-balancing and the confidentiality of the inside information for the route computation and the path provisioning among different carriers. We also provide measured performance results for inter-domain path computation and path setups using global optimization scheme, estimating the path provisioning time at the network scale dependency with the route computation time and the signaling time. As a result, we show its applicability to a 1000-node scale photonic network.

Optical network control challenges

This paper discusses two challenges, failure recovery enhancement and scalability enhancement, on optical network control plane. These challenges enable us to provide optical networks with high reliability and lower operational cost.

Multi-layer and Multi-domain Network Orchestration and Provision of Virtual Views to Users

Recently, there are expectations for transport networks among wide area to provide paths dynamically applying Software-Defined Networking (SDN) for faster provisioning paths with low latency and high availability. However, transport networks are actually heterogeneous networks consisting of multiple layers and domains unlike data center networks where SDN have been applied. In addition, virtualization of network resources is also expected by applying SDN. We propose a network orchestrator based on ODENOS for orchestrating multi-layer and multi-domain networks and for providing virtual views to users. Proposed network orchestrator has four features. Firstly, a common data model is introduced for solving a problem of differences among layers and domains. Secondly, relations between layers/domains are managed for solving a problem about lack of managing relations between layers/domains. Thirdly, data model and interfaces are converted between the network orchestrator and layer/domain for solving a problem of differences of methods of path setup among layers/domains. Lastly, a virtual view is created for each user with separated topology based on an ID of each user for solving a problem about lack of a method to create separated views for each user. Results of performance evaluations show that the proposed network orchestrator can dynamically provide virtual views to users on demand.

Experimental performance evaluation of inter-domain path provisioning with multiple PCEs

In this paper, PCE-based inter-domain path computation has been discussed. Interworking of multiple PCEs with the BRPC procedure enables not only end-to-end shortest path computation but also end-to-end diverse path computation over multiple networks. We have analyzed various PCE-based inter-domain path computation strategies using BRPC, in terms of the path computation complexity. To reduce path computation time, a simplified path computation strategy was proposed and evaluated experimentally with our prototyped PCE system. Our experimental results show good correspondence with analytical ones. The proposed strategy drastically reduced path computation time down to one-seventh in comparison with the conventional strategy.