Online proceedings

I. INTRODUCTION

Large capacity and economical optical networks need to have intelligent functions such as dynamic resource allocation and efficient resource management: the best known example is GMPLS (Generalized Multi Protocol Label Switching). GMPLS requires IP to transport the signaling and routing protocols over control plane. Therefore, to implement the GMPLS network, it is necessary to set component parameters for nodes and links and protocol operation parameters for GMPLS protocols. This is so expensive that plug and play techniques to set these parameters are needed.

II. PLUG AND PLAY TECHNIQUES

We propose plug and play techniques with automatic parameter setting (Technique 1) and exchanging (Technique 2) for the component parameter configuration. The overview of the plug and play techniques is shown in Fig. 1.

Fig.1 Overview of Plug and Play techniques

A. Auto configuration setting for local node’s parameters (Technique 1)

Technique 1 enables GMPLS component parameters of a local node to be set upon triggered by a link-up (optical cable connection). The parameters are generated by the combination of the previously assigned node identifier and unnumbered link identifier. Therefore, the parameters are unique in the network. For applying the proposed technique to all optical equipments, we propose all optical node architecture with a passive component and an active component for link-up search. As shown Fig. 1, our proposed optical node has a N x N all optical switch element as the passive element and a trunk element as the active element. In this case, the proposed optical nodes connect the switch port, which is connecting to trunk element, to other switch port randomly for the link-up search. The trunk element detects a link-up if the trunk port connects to ports lead to the neighbor node and sends the notification of link-up to a control element. When the control element receives the notification, it generates the component parameters of the local node.

B. Auto configuration exchanging for remote node’s parameters (Technique 2)

Technique 2 enables GMPLS component parameters of a remote node to be set by message exchange between neighbor nodes. As the message, we use messages from an extended version of LMP. The message is includes the local node component parameters generated by Technique 1. The config message is used as the exchange message for a control plane interface and the test message is used for a data plane interface. Therefore, the control element identifies whether the link-up port is a control plane interface or a data plane interface.

III. IMPLEMENTATION AND EVALUATION

We implemented our proposed techniques in all optical layer-1 nodes and evaluated the setup time of network configurations in an evaluation network composed of three layer-1 nodes. For the total twelve interface ports kept by three layer-1 nodes, it took about 18 seconds to set up component parameters of them. The proposed technique automatically can generate all link parameters including control channel link parameters and data channel link parameters and reduce the number of operation man-hours needed to establish the service. Fig. 2 shows a demonstration of our techniques. We expect that LMP extension for our proposed plug and play technique become a standard.

Fig.2 Demonstration of our proposed techniques

GMPLS is the unified control plane architecture for connection management over various data plane technologies. It consists of a set of protocols: OSPF/ISIS-TE, RSVP-TE and LMP; which together reduce manual configuration by automating provisioning, while still supporting centralized network management. GMPLS technology adaptation is mostly concerned with the cross-connection of data plane resources, i.e., the definition of the generalized forwarding label and related procedures. However, there are additional data plane functionalities that would benefit from automated single-sided provisioning.

This talk will focus on how GMPLS can be used to automate the establishment and configuration of Operations and Management (OAM) functions associated to connections.

OAM is essential to ensure that the desired service level of traffic engineered connections is met. In certain technologies OAM entities are inherently established once the connection is set up. However other technologies, especially OAM for packet switched networks (e.g., Ethernet PBB-TE, MPLS(-TP)), require an extra configuration step after connection establishment to setup OAM entities. It would be desirable to bind OAM setup to connection establishment signaling to avoid two separate management/configuration steps (connection setup followed by OAM configuration) which increases delay, processing and more importantly may be prone to miss-configuration errors.

In some situations the use of OAM functions, like those of Fault- and Performance Management, may be optional depending on actual network management policies. Hence the network operator must be able to choose which kind of OAM functions to apply to specific connections and with what parameters the selected OAM functions should be configured and operated. To achieve this objective OAM entities and specific functions must be selectively configurable. OAM functions are technology specific, i.e., vary depending on the data plane technology. In addition for a given data plane technology a set of OAM solutions may be applicable.

The GMPLS OAM configuration framework, introduced in this talk, will allow the selection of a specific OAM solution to be used for the signaled LSP and provides technology specific information for detailed configurations.

With the Internet traffic growing exponentially, the current backbone network faces serious scaling and economic issues. In this presentation, we propose an approach named HoTNet (Holistic Transport Network) to deal with above issues. The basic idea of HoTNet is introducing the MPLS-TE fully meshed or OTN fully meshed architecture to build backbone network, using the multi-layer planning tool to optimize network resources statically, using the Bypass Server, Path Computation Element (PCE) and Generalized Multi-Protocol Label Switching (GMPLS) to optimize network resources dynamically. Theoretical analysis and network simulation proved the HoTNet solution can bring TCO saving up to 40% for carriers in different network architectures.

The prototype we implemented in test bed is based on emulate routers and commercial OTN equipments, and introduced some new network elements such as Bypass Server (BS) and Path Computation Element (PCE) to make bypass decision and execution. With this prototype, we can demonstrate the whole traffic offloading procedure which includes traffic monitoring, bypass decision, bypass link routing and signaling in different layer coordinately, and service switching with make-before-break. Through the automatic traffic offloading, carriers can cope with the dramatically increased traffic by an OpEx effective way. The following figure shows a simplified topology of HoTNet test bed:

Today, PCE-based routing model is believed to be the most suitable for path routing in the multi-domain ASON/GMPLS networks. At the same time, with the increment of applications for home, enterprise and science computation, it is required for PCE to support more advanced services, such as multicast, VPN and in-advance reservation etc. However, the extension to GMPLS protocols is not mature to support such advanced services.

In this presentation, we discuss the framework for implementation of Web Services (WS) based PCE. Web Services is a software architecture that supports interoperable machine-to-machine interaction in network. In this architecture, services that will be required by upper or peer applications are advertised with the description in Web Service Description Language (WSDL). The Simple Object Access Protocol (SOAP) and Extensible Markup Language (XML) are taken for the machine-to-machine communications. Many network interoperation architectures [1][2] have adopted Web Services in Optical Networks as it would bring several advantages:

• All functions are defined as services with universal interfaces. Details of internal implementation have been concealed.
• Both WSDL and XML provide flexibility. WSDL could describe new services in a machine readable way. XML could supports extension in inter-machine communication easily.
• Standardized mechanisms for user authentication and policy management.

• PCCs require services by interacting with interfaces provided by WS-Gateway.
• Cooperation between WS-PCEs for inter-domain paths computation.
• WS-TEDs collect intra-domain Traffic Engineering (TE) information.
• Cooperation between WS-TED for inter-domain TE information advertising.

Fig. 1 Architecture of the Web Service based PCE

Fig. 2 Topology of 3TNet testbed

We have demonstrated inter-domain multicast, OVPN and in-advance reservation services mentioned above based on the Web Service based PCE in 3TNet testbed [3][4]. Fig. 2 shows the optical network layer topology of the field trial network which is carried out in Yangtze River Delta. There are 13ASON nodes located in two domains. In the data plane, all the ASON nodes are SDH cross-connects provided by Huawei’s (ONS 9500) and Fiberhome’s (FonsWeaver 780) with at least 640 Gb/s non-blocking switching capacity. The inter-networking point is located in the SJTU network center. 2 GbE and 2 STM-16 links connect two domains.

Reference:
[1]J. Lehman et al., “Control Plane Architecture and Design Considerations for Multi-Service, Multi-Layer, Multi- Domain Hybrid Networks”, High-Speed Networks Workshop, INFOCOM 2007.
[2] B. St. Arnaud et al., “Customer-controlled and ?managed optical networks,” J. Lightw. Technol., vol. 21, no. 11, pp. 2804?2810, Nov. 2003.
[3]J. Wang et al., “Field Trial of Inter-domain Point-to-multipoint Connections in ASON Using Web Service Mechanism”, OFC 2007.
[4]Y. Wang et al., “SOA-Based Inter-Domain OVPN Service for Coordinated Scheduling of Distributed Computing”, OFC 2008.

As increasing Ethernet service users and enhancing service functions, a next generation wide area layer2 network beyond the state-of-the-art carrier grade Ethernet is expected. This presentation shows GMPLS controlled Next Generation Layer2 (NGL2) network. The presentation also shows the results of the experiment regarding with establishing a path in NGL2 network by RSVP-TE.

Figure 1 shows the architecture of our proposed NGL2 network. This network consists of a provider domain and user domains. User Network Interface (UNI) supports various kinds of Ethernet frame and provides connection oriented Point-to-Point/Point-to-Multipoint layer2 path service by setting up VLAN in response to users’ requests. In the provider domain, User MAC frame from the user domain encapsulated into NGL2 MAC frame shown as Figure 2. The layer2 path is identified by VLAN ID of NGL2 MAC frame. In the conventional VLAN, the same VLAN ID is used through the layer2 path. In NGL2, VLAN ID is swapped link by link for ensuring the scalability. So, VLAN ID is used as Virtual Path Identifier (VPI), and each layer2 switch has a VPI convert table. Because the management of VPIs becomes complex due to swapping, a layer2 path is established by using GMPLS as Control Plane.

There are two main challenges to establish a layer2 path by GMPLS. The first one is the advertisement of Destination MAC address (D-MAC) and Source MAC address (S-MAC) of NGL2 MAC frame. The second one is the dynamic configuration of VPI convert tables. In this research, RSVP-TE is extended to achieve these two challenges. For the first challenge, D-MAC and S-MAC in NGL2 are created by using Router IDs and Interface IDs in Control Plane. In this way, it is not necessary to advertise these MAC addresses newly. For the second challenge, VPI convert tables are configured by RESV message. Also, to inform end nodes about the information of VPIs, the values of labels assigned for the established layer2 path are written into Record Route Object of RESV message.

The experiment to establish a layer2 path by the extended RSVP-TE is done. NGL2 prototype system is implemented. This system provides In-band Message Communication Channel (IMCC) for Control Plane and MAC-in-MAC transport mechanism for Data Plane. It is confirmed that NGL2 prototype systems are configured by the extended RSVP-TE, data can be exchanged between user PCs, and D-MAC and S-MAC of NGL2 MAC frame is set up dynamically and correctly.

Figure 1: Architecture of our proposed NGL2 network.

Figure 2: NGL2 MAC frame format.

GMPLS network technology has been developed mainly for the time-division multiplex (TDM), fiber switch capable (FSC). However, considering the rapid growth of Ethernet services these days, the Ethernet control technology based on layer-2 switch capable (L2SC) becomes quite important. In addition, inter-carrier L2SC should be also developed for the flexible and efficient control of wide-area/global Ethernet services. The L2SC for the single domain and between multiple GMPLS domains has been developed, but the inter-carrier L2SC connection between ASON and GMPLS domains has not been investigated. From this viewpoint, this presentation describes the implementation of L2SC control between ASON and GMPLS domains. This presentation also provides the experimental result of interoperability test of L2SC between ASON and GMPLS domains.

As the issues of the inter-domain routing between ASON and GMPLS networks, firstly this presentation describes the difference of the definition regarding the destination node address between the ASON and GMPLS networks. Nextly the routing functionalities in order to exchange and advertise the reachability information of destination node address between the domains are shown. As for the signaling, this presentation notes the different types of explicit route object format between ASON network with multi-session signaling and GMPLS network with single-session signaling. Then the RSVP-TE signaling functionalities on the intermediate AS border router (ASBR) are described so as to accept these types of route description and calculate the route toward destination or next ASBR. In addition, the implementation in order to accept the differences of RSVP-TE object between ASON and GMPLS network is also described. In this L2SC implementation, the fixed port-label value consisting of the tagged/untagged flag, physical port and VLAN-ID is utilized. However, considering the effective utilization of VLAN-IDs in inter-domain Ethernet connection, a VLAN-ID should be selected from available multiple VLAN-IDs assigned to single TE-link. So this presentation discusses effective utilization of this information.

Finally this presentation is summarized by providing brief overview of L2SC LSP creation experiment between ASON and GMPLS in the Keihanna Interoperability Project.

Introduction

Testbed setup

Testing is being performed in a testbed consisting of a number of virtualized PCs running GNU/Linux and a homebrewn user-space MPLS implementation. The MPLS software is able to encapsulate Ethernet frames and forward them in the standard MPLS manner, with the difference that G-ACH support has been added. By using Ethernet over IP tunnels different topologies can easily be created and used to run test scenarios that utilize the G-ACH.

Test cases

Four test cases have been identified for prototyping:

• single LSP. This is the basic case to show how MPLS-TP forwarding should be done, with a focus on G-ACH and GAL processing and the use of TTL field to address an MIP in the middle of an LSP. See figure 1.
• Nested LSPs. This case will investigate how TTL field should be handled when nested maintenance association levels are used. See figure 2.
• Nested LSPs with working and backup path over the same number of hops. This case will investigate how a failure on the working path can be detected using MPLS-TP OAM and how this can trigger protection switching. See figure 3.
• Nested LSPs with working and backup path over different number of hops. This case will investigate how the GAL node addressing will be influenced by the hop difference in a protection switching scenario. See figure 4.

Results and conclusions

The prototype result will provide a primary understanding of how MPLS-TP OAM should be implemented, identify limitations of the current OAM specifications, provide OAM solutions for the pre-defined scenarios and propose necessary improvements to the ongoing MPLS-TP standardization work.

¹ As proposed in the current Internet-Draft “MPLS Generic Associated Channel”.

Figure 1: Test case 1, a simple LSP.

Figure 2. Test case 2, LSP nesting.

Figure 3. Test case 3, protection switching with equal length paths.

Figure 4. Test case 4, protection switching with unequal length of paths.

We propose the next generation packet transport system, which partially uses the MPLS-TP (Multi-Protocol Label Switching-Transport Profile) technologies. The concepts of the system are (1) High reliability using OAM (Operation Administration and Maintenance) and protection, (2) Assured packet transfer for each users using static LSP (Label switched Path) control with NMS (Network Management System). In this presentation, High reliability issues are discussed, and in the other presentation we discuss about assured transfer issues.

1. Backgrounds
As network application used in business scene, importance of networking for business is expanding. For such networks, by now, two kinds of network are used. One is leased line, for example, ATM (Asynchronous Transfer Mode) access services, which is very reliable but relatively slow. The other is packet-based VPN (Virtual Private Network) services, for example, IP-VPN or Metro Ethernet, which is very fast and best-effort basis.
As the importance of networking increases, the network is required both wide band-width and reliability. The valuable length packet technologies, for example, Ethernet or IP forwarding progressed nearly, it is advantageous to use these technologies to make high speed network. On the other hand, it is very crucial to make networks reliable, so that they maintain their connectivity as much as possible. To meet these requirements, we tried to develop packet-based reliable network.

2. Development of MPLS based packet transport system
For that purpose, we developed the MPLS-TP based packet transport system. To proceed the manageability of the network, we use the static MPLS LSPs which are managed from NMS (Figure 1). Each LSP is supervised with embedded OAM tools. Moreover when one LSP is set on the network, NMS checks the capacity of resources on each transport system, then assured transfer is accomplished.

3. Implementation of OAM technologies on the system
It supports the OAM functionality for each LSPs (users), continuity verification, AIS/RDI, and Loopback. As all the technologies are implemented as hardware, these functions can be performed respectively. For all the LSPs, 1:1 protection with APS (Automatic Protection Switching) mechanism is implemented.

4. Conclusion
We developed the next generation packet transport system, which can be used to make very reliable network. In the system, we used static path with MPLS label set by NMS and supervise these paths with embedded OAM tools.

Figure 1. Proposed packet transfer system

The sustaining growth of data traffic volume calls for an introduction of an efficient and scalable transport platform for links of 100 Gb/s and beyond in the future optical network. In this presentation, we first discuss the existing major scalable optical network technology options and present their strengths as well as weaknesses. Next, we describe the concept and architecture of a novel, spectrum-efficient and scalable optical transport network architecture called spectrum-sliced “elastic optical path” network (SLICE). SLICE brings numerous enhancements in spectral-efficiency and new degree of spectrum assignment in the form of:

• Efficient and scalable accommodation of sub-wavelength, super-wavelength, and multiple rate data traffic
• Elastic bandwidth variation for efficient utilization of network resources and enhanced survivability
• Future-proof, efficient accommodation of future data rates and modulation formats.

Dynamic bandwidth variation of elastic optical paths provides network operators with new business opportunities offering cost-effective and highly available connectivity services through time-dependent bandwidth sharing, energy-efficient network operation, and highly survivable restoration with bandwidth squeezing. As the enabling technologies of SLICE concept, we discuss an optical orthogonal frequency-division multiplexing-based flexible-rate transponder and a bandwidth-variable wavelength cross-connect. Then, we present technical challenges including management and control issues that arise in this new network architecture. We believe that SLICE will introduce a new degree of freedom and open up a more significant role for optics in future transport networks.

Ericsson has recently realized a test bed in its site in Coventry implementing a GMPLS stack for the control of WSON networks. Purpose of this installation is the Proof of Concept (PoC) of a control plane controlled photonic network. Such implementation of WSON networks is based on the interaction between three different entities: a Network Management Application acting also as a PCE, the NE and an external Linux PC hosting the Control Plane for each of the NEs. The network elements used in this test bed are commercial ROADMs (MHL3000) with four ways WSS and the network is made by five NEs.
The following features have been demonstrated:

• Auto-Provisioning
  • Auto-provisioning with path definition (unprotected circuit)
  • Auto-provisioning with restoration (restorable circuit)
• Restoration
  • All the circuit paths are pre-calculated by a centralized PCE that knows in advance both the traffic matrix and the network impairments
  • Pre-planned restoration
    • Single fault restoration
    • Double fault restoration
  • Optical restoration with card sharing
• Network auto-discovery
  • Network topology and configuration upload

Once that the LSPs are signaled the network is fully controlled by the control plane, that is restoration is done directly by NEs without any intervention either from human operator or NMS.

The WSON stack includes the entire set of GMPLS protocols: OSPF-TE as routing protocol, RSVP-TE as signaling protocol and LMP for the management of the links. Due to the fact that the PCE runs offline, OSPF-TE is not used to feed the PCE database but is used by the different NEs to validate the reserved path to avoid signaling LSP along no more available paths. All the protocols have been implemented following IETF related standards.
The presentation will show in details the test bed environment and the major result of this PoC.

The Path Computation Element (PCE) has become a familiar concept and multiple implementations exist to determine end-to-end paths in multi-domain MPLS-TE and GMPLS networks. PCEs can be used in per-domain path computation techniques to devolve the computation of a path segment to each domain entry point. Alternatively, the backward recursive path computation (BRPC) mechanism allows the PCEs to collaborate to select an optimal end-to-end path that crosses multiple domains.

These techniques, however, assume that the sequence of domains to be crossed from source to destination is well known. No explanation is given of how this sequence is generated or what criteria may be used for the selection between domains. In small clusters of domains, such as simple cooperation between adjacent ISPs, this selection process is not complex. In more advanced deployments (such as optical networks constructed from multiple sub-domains, or multi-AS environments) the choice of domains in the end-to-end domain sequence can be critical to the determination of an optimum end-to-end path.

In this presentation we explain the latest mechanisms for the selection of domains that comprise the end-to-end domain sequence. We examine how BRPC could provide the necessary function, but show how it doesn’t scale to large networks. Instead, a new model of hierarchical PCEs is explained. This way of performing PCE cooperation enables rapid and scalable selection of optimal paths and can take advantage of both typical traffic engineering constraints (hop count, bandwidth, lambda continuity, path cost, etc.) and also more commercially relevant constraints such as policy, SLAs, security, peering preferences, and dollar costs.

In the past few years, ASON/GMPLS control plane has been deployed in service provider’s commercial networks, and their advantage in terms of operational cost (OPEX) reduction was reported [1]. However, most of such networks are still relatively small, which are constructed by few tens of nodes, and they have to be expanded to nation-wide larger networks for more effective OPEX reduction. To control larger networks, typically thousands of nodes, we have studied qualitatively and concluded that the PCE-based routing was the most suitable model rather than other routing models, the per-domain routing and the ASON hierarchical routing [2]. But quantitative study using a large network test environment has not been reported yet.

The purpose of the study is to demonstrate, quantitatively, scalable PCE+GMPLS control architecture, in large size of multi-domain networks. In order to evaluate its scalability, a PC-based 1000-node test network has been developed in our Lab. This test network is composed of twelve service domains, and each of which has several OSPF areas. GMPLS controllers (i.e. OSPF-TE and RSVP-TE) are implemented in the PC-based nodes, and a PCE is located in each area. To compute inter-domain paths, PCEs support BRPC for shortest path computation as well as an automatic PCE selection mechanism based on routing information of control plane networks. Also PCEs and GMPLS controllers support Path Key, so as to setup end-to-end path with preserving confidentiality among domains.

In the presentation, we will discuss impacts to PCE+GMPLS control plane when the number of service domains or areas is increasing, and will also provide measured performance results for inter-domain path computations and path setups.

Acknowledgement
This work is a part of “Lambda Utility Project” supported by National Institute of Information and Communications Technology (NICT).

Reference
[1] S. Liu and L. Chen, “Deployment of Carrier-Grade Bandwidth-on-Demand Services over Optical Transport Networks: A Verizon Experience”, OFC/NFOEC 2007 NThC3, 2007.
[2] I. Nishioka, S. Ishida, Y. Iizawa, “End-to-End path routing with PCEs in multi-domain GMPLS networks”, iPOP2008 T3-2, 2008.

Multicast services are increasingly in demand for high-capacity applications such as multicast VPNs, IPTV (on-demand or streaming), and content-rich media distribution (for example, software distribution, financial streaming, or data-sharing).

Operators of packet networks have been quick to recognise the benefits of multicast IP and point-tomultipoint MPLS-TE as tools to optimise network resource usage, and to transmit only one copy of data on shared links towards the multiple destinations. Additionally, the use of MPLS-TE helps provide Quality of Service through resource reservation, traffic engineering, and the management of back-up paths.

However, in a multi-layer network the benefits of IP multicast and P2MP MPLS-TE cannot be fully realised without also considering resource sharing in the optical transport layer. For example, as shown in Figure 1, a router may consider itself to be immediately adjacent to two down stream routers in the distribution tree meaning that it should make copies of the data and transmit them separately. But within the optical network, the paths between the routers may be substantially shared. The most optimal use of the network can only be achieved by utilising P2MP in the optical layer.

Figure 1: Multi-layer P2MP

This presentation examines how a multi-layer P2MP solution can be built, and looks at the implications for P2MP in multi-domain networks. We examine how multi-domain issues in the packet layer are more prevalent in the transport layer, and show how to combine network planning techniques with new Path Computation Element ideas to achieve optimal delivery of high-capacity multicast services over multidomain optical networks.