講演予稿集

Thursday 5, June 2008 10:00-12:20

Opening address
- Tomonori Aoyama, General Chair, Keio University, Japan
- Bijan Jabbari, General Chair, ISOCORE, USA

Keynote
K-1 "Optical Network: from FTTH to Transparent Backbone Network"
- Hiromichi Shinohara, NTT, Japan

Biography: Mr. Hiromichi Shinohara has been a Vice President of NTT, Executive Director of NTT Information Sharing Laboratory Group since June 2007.
He joined NTT Laboratories in 1978. He has consistently been spending his carrier to realize FTTH. In addition, he has recently been engaged in strategic planning and promoting of research and development for NGN architecture and platform technologies.

There are several proposed changes to the transmission/transport infrastructure, including the replacement of SDH equipment, the use of "packet transport", and tighter integration of packet switching and optical equipment. Along with these proposals come suggestions for new protocols to achieve these changes, such as PBB-TE, T-MPLS, "plain old MPLS", and GMPLS.

In this talk, I will review some of the drivers for change, and take a big step back to look at the big picture of how we got here and where we should go. I will conclude with a proposal that I believe essential to "Next Generation Networks" and efficient and cost-effective network infrastructure. I will try to outline the impact that such an approach will have on Service Providers and equipment vendors of both switching and transmission gear.

Program introduction
- Kireeti Kompella, Juniper Networks, USA and Kohei Shiomoto, NTT, Japan

Exhibition introduction
- Takeshi Akaike, NTT and Satoru Okamoto, Keio University, Japan

Tech. Session 1: ASON
Thursday 5, June 2008 13:45-15:00

Chair: Kireeti Kompella, Juniper

T1-1 "Control Plane Resilience and Security in GMPLS Networks : Fact and Fiction"
Adrian Farrel, Old Dog Consulting, UK

Optical transport network security has not traditionally been a topic of great concern to service providers. The only openings for attack on a transport infrastructure have been either through the management plane or by physical intrusion either at the equipment site or on the fibre links between sites. Techniques such as IP access lists and passwords have been considered adequate to protect devices from access through private management networks.

The introduction of the GMPLS control plane has exposed devices to a greater range of message sources and potential commands that can change the configuration of network equipment. But the DCN has still been considered to be a private network with the result that no stronger security techniques have been deployed.

This talk will explain how this naive view of the GMPLS DCN results in confusion about control channel protection and restoration, and could lead to major vulnerabilities in core networks that compromise service delivery.

The speaker will describe how utilizing an IP-based control plane means more than just assigning IP addresses to the ends of dedicated channels such as OSCs. He will show how the definition of a GMPLS control channel in RFC 4204 expresses a logical entity that is independent of the underlying physical connectivity. As a result, discussions of backup control channels and control channel recovery techniques within GMPLS are bogus and should be largely restricted to physical schemes similar to link level protection.

But a consequence of the introduction of the GMPLS control plane is that the network is more open to security attacks. This talk will highlight those vulnerabilities, explain how the GMPLS protocols protect themselves from attack, and explain the areas in which there is still scope for intrusion.

Biography: Adrian Farrel is co-chair of the IETF's CCAMP, PCE, and L1VPN working groups. He is author of GMPLS : Architecture and Application and The Internet and Its Protocols, and is co-editor of the new book MPLS : The Next Steps. Adrian runs a thriving consultancy, Old Dog Consulting, specialising in GMPLS, MPLS, PCE, and routing protocols.

T1-2 "e2e GMPLS P&R with next generation interface for the green networking."
Hidetsugu Sugiyama, Juniper Networks, Japan

All optical switching transport networks can reduce the power consumption of the whole network system to solve the issue of "global warming" by reducing “OEO switching node” in transport network. As the network system can reduce the power consumption, we call such the networks "green networking". In order to build the "green networking", all circuit switching nodes should support the "photonic cross connect (PXC/OOO)" as it is not necessary for packet buffering. In other words, only packet switching nodes should remain as “OEO switching”.

However, there is a known issue that OOO transport node can not provide fast link protection as the optical is terminated at packet switching node such as a Router who sends or receives the data packets on single wavelength in the optical fiber.
Therefore, all optical switching transport networks are still not realistic for deployment if users need high availability circuit network. This situation is based on legacy router interface that can support only single wavelength in the optical fiber and connect to transponder in optical transport node.
If router data port has a capability to control dual wavelengths (or multiple wavelengths) in the optical fiber, this circuit protection issue will be solved as router can send or receive data packets by selecting one of available wavelengths in the data port.

In this presentation, “GMPLS based e2e protection” mechanism on new router interface called “Tunable OTN interface” is introduced, by comparing other protection mechanism such as MPLS FRR, etc.. And also we will discuss several issues and solutions in terms of technical point of view, in terms of operation point of view, and in terms of business point of view.

Biography: Hidetsugu Sugiyama is an R&D Support Director, Technical Operations APAC at Juniper Networks and manages Open IP Solution Development Program in Asia Pacific.
He is currently focused on next generation IP Optical network integration technology relating MPLS/GMPLS products.　He has over 19 years experience in Telecommunication industry as a Network consultant engineer in data communications.

T1-3"Automated End-to-end Circuit Provisioning for high bandwidth demand applications: the challenges and our experiences in 3TNet"
Weiqiang Sun, Guowu Xie, Yaohui Jin, Wei Guo and Weisheng Hu, Shanghai Jiao Tong University, P. R. China

Optical networks, with the tremendous bandwidth it can provide, have been regarded as a premium choice for connecting various systems with high bandwidth demand. Traditionally, optical networks have been used mostly in the form of dedicated links, or in connecting a relatively small number of systems. While the demand for high bandwidth increases in intensity over the years, its extensity has also grown significantly. Optical networks in its historical form with static configurability and coarse granularity is no longer transport candidates for emerging applications, such as E-Science, storage area networking and grid computing.

GMPLS is an important feature added to existing optical networks. It can automate the circuit provisioning process hence simplify network OAM and can potentially reduces OPEX. At the same time, it opens up opportunities for applications that have high bandwidth demand and also require high levels of re-configurability and flexibility. Numerous efforts have been witnessed in the last 5 years, in applying GMPLS enabled optical networks to high bandwidth demand applications. But two questions are still not answered. First, traffic generated by applications in general is much more dynamic then those aggregated in a Metro core node. With the traffic changes quickly with time and significantly in volume, will the network with distributed control plane be able to adapt to it, with acceptable efficiency and stability? Second, end systems generally has inherent multi-point communication requirement. How can an optical network with the tradition of provisioning dedicated circuits meet such a requirement, and in a scalable way?

In this presentation, we will try to present our experiences in dealing these two challenges in a regional testbed in China - 3TNet. First, we will introduce our intensive testing over the testbed regarding the provisioning performance of a typical GMPLS enabled network. We find that the provisioning performance can vary significantly between vendors and under different traffic patterns. We have concluded our findings in an Internet Draft submitted to IETF CCAMP working group [1]. To our knowledge, this is the first draft that tries to characterize the connection provisioning performance of GMPLS networks.

Second, we will present our experiences in building a sophisticated operation support system (OSS), over a GMPLS network. In the OSS, we implemented a proprietary application programming interface (API) on top of the control plane, such that the network resources can be invoked directly from VoD and IPTV services. We also introduced a VLAN based technique to increase the connectivity of end system interfaces. This feature will greatly increase the flexibility of end system circuit provisioning [2].

Reference
[1]W. Sun et al., Label Switched Path (LSP) Dynamical Provisioning Performance Metrics in Generalized MPLS Networks, www.ietf.org/internet-drafts/draft-xie-ccamp-lsp-dppm-02.txt, Nov. 2007
[2]Weiqiang Sun et al., A cross-layer optical circuit provisioning framework for data intensive IP end hosts, IEEE Communications Magazine, Vol.46, No.2, pp.S30-37, Feb. 2008

Biography: Weiqiang Sun is an assistant professor in the Department of Electronic Engineering at Shanghai Jiao Tong University (SJTU), China. He received his B.Tech. from the Special Class for the Gifted Young (SCGY) at the University of Science and Technology of China (USTC) in 1999 and his Ph.D. from the same university in 2004. His research interests include dynamically configured optical networks, QoS in packet-switched networks, and IPTV.
Dr. Sun has served as invited speaker on AOE 2005 (Shanghai, China), COIN-NGNCON 2006 (Jeju, Korea) and ANTS 2007 (Bombay, India). He is a member of Technical Program Committee of COIN 2008 (Tokyo, Japan). He has one IETF draft contribution in CCAMP working group and more than 40 peer reviewed papers on International Journals and Conferences.

@Convention Hall
Thursday 5, June 2008 15:15-16:30

Chair: Ernesto Damiani, University of Milan

P-1 "Optical Fast Reroute"
Adrian Farrel, Old Dog Consulting, UK

Fast Reroute (FRR) is a familiar and established technique in MPLS networks where Label Switched Paths (LSPs) are can be rapidly routed around failures by tunneling them through protection tunnels. FRR allows end-to-end LSPs to be repaired without waiting for routing protocols to converge, and without requiring end-to-end protection. New techniques are also being developed for Fast Reroute in IP networks (IP-FRR) to rapidly repair end-to-end delivery without waiting for routing protocols to converge.

MPLS FRR was developed for use in packet networks and relies on the use of label stacking, that is, the nesting of one LSP within another. Although hierarchical LSPs are possible in optical networks, they require a mix of technology layers or sub-layers and so are not suitable for use to provide rapid repair in most optical networks.

This presentation will examine the latest standardization efforts within the IETF that give rise to new options for rapid local repair in optical networks.

• Segment protection is a technique that allows planned protection spans to be signaled as part of the initial LSP setup. It offers dynamic control plane management of service delivery that closely matches the techniques that have traditionally been manually configured in transport networks.
• LSP stitching combines the control plane behavior of hierarchical LSPs with the data plane characteristics of a single end-to-end LSP. Developed to enable inter-domain LSPs in packet networks, LSP stitching turns out to be very useful in optical networks where LSP nesting is not possible.

In this talk, the speaker will show how the combination of segment protection and LSP stitching can provide many of the benefits of Fast Reroute to provide flexible and dynamic fault protection within optical networks.

Biography: Adrian Farrel is co-chair of the IETF's CCAMP, PCE, and L1VPN working groups. He is author of GMPLS : Architecture and Application and The Internet and Its Protocols, and is co-editor of the new book MPLS : The Next Steps. Adrian runs a thriving consultancy, Old Dog Consulting, specialising in GMPLS, MPLS, PCE, and routing protocols.

P-2 "GMPLS VLAN Path Establishment using Inter-domain VLAN Tag Swapping"
Kou Kikuta, Masahiro Nishida, Daisuke Ishii, Satoru Okamoto, Naoaki Yamanaka, KEIO University, Japan

Wide-Area-Ethernet Service provides multiple inter-connections among customers’ LANs. In Wide-Area-Ethernet network, Virtual paths are established on layer-2 by VLAN (Virtual LAN) technology. Network switches read VLAN tag, path identification extended bit in Ethernet frame, and determine forwarding direction for each frame. We need to setup VLAN configuration of each switch to establish new path.
GMPLS is able to setup the switches’ configuration automatically and establish the new path just after customer demand. We are researching on this protocol’s details, exchanged VLAN information, controlling switches, using limited numbers of VLAN IDs, and others.
We presentation addresses Layer-2 Switching Capability (L2SC) with VLAN tag swapping. In Wide-Area-Ethernet network, same VLAN ID is used throughout a path. However, a number of VLAN ID is limited to 4096 due to extended frame format, thus there is a limit to the number of paths. VLAN swapping enables to use more VLAN IDs in a network like a Wavelength Converter.
We implemented a VLAN swapping function on Linux based PC and controlled the function by using GMPLS. In the experimental network, we used VLAN tag swapping switches that implemented by Linux PC. Multiple paths can be established by VLAN tag based switching that includes VLAN tag swapping.

Biography: Kou Kikuta was born in Kanagawa, Japan on July 5, 1984. He received B.S degree from Department of Applied Chemistry, Keio University, Japan in 2007. Currently, he is a master course student of the school of Science for OPEN and Environmental Systems, Keio University. His current research interest is GMPLS protocols and next generation layer-2 network technologies.

P-3 "Chromatic Dispersion Compensation Control in Dynamically Reconfigurable All-Optical Networks"
Eiichi Horiuchi, Misato Kamei, Shoichiro Seno, and Yoshimasa Baba, Mitsubishi Electric Corporation, Japan

Advances in optical technology have realized all-optical DWDM networks where wavelengths are switched and transferred without OEO conversion. Dynamically reconfigurable mesh topology networks can be constructed by deploying all-optical switching nodes such as ROADM and PXC. ASON/GMPLS technology enables automatic configuration, dynamic path setup, and fault recovery such as shared mesh restoration for optical networks. For the all-optical networks, functional extension of the control technology is required to cope with optical signal impairments and other routing constraints such as switching capability. The impairments are not only routing constraints but may also require compensations on-the-fly during a lifetime of a dynamic wavelength path. Chromatic dispersion is the dominant factor of the impairments for regional and domestic backbone networks. This presentation focuses on the chromatic dispersion compensation, and proposes control means of measurement, management, and compensation in the dynamically reconfigurable all-optical networks.

Although various techniques of measurement and dynamic compensation for chromatic dispersion have been proposed, requirements and problems for control plane are in general as follows.
1) Measurement function is implemented in every node in the network and measurement is executed automatically. It requires management of resources such as wavelengths and the measurement function.
2) Before a wavelength path setup, dispersion value is considered as a routing constraint e.g. compensation capability is limited.
3) For high bit rate over 40 Gb/s, tolerable dispersion is very small. Therefore dispersion accumulated over a wavelength path should be measured or estimated accurately to configure compensation device(s) along the path. Factors of residual dispersion after compensation are A) changes of configuration or environment such as exchange of fiber and fluctuation of temperature, B) errors of estimation or measurement such as dispersion slope and resolution.
4) Time to make a connection with sufficient quality of the optical signal should be minimized especially during fault recovery. Accurate dispersion to be compensated over the path should be quickly configured on compensation device(s). If the configured dispersion value is inaccurate, additional adjustment of the compensation takes much time to reach sufficient quality.

Dispersion measurement can be conducted per-span (link) or per-path automatically and periodically. The measurement per-span is easier to implement, while compensation with summation of spans along a path is not always available. This is because errors due to resolution of per-span measurement and dispersion of devices in nodes accumulate along the path, and the residual dispersion after compensation may exceed tolerable dispersion as a result. The measurement per-path is required for such case. However, it has constraints: 1) It can not be conducted before a wavelength and a route for the path is determined 2) Resource contention occurs between different measurements.
By taking into account the above issues, we propose GMPLS extensions which control dispersion measurement per-span and per-path, and control dynamic dispersion compensation for dynamically routed paths (Figure 1).

Fig.1 Control of Measurement and Dynamic Dispersion Compensation by GMPLS extension

Biography: Eiichi Horiuchi received the B.E. degree in electronics engineering from Osaka University, Toyonaka, Japan in 1990.
In 1990, he joined Communications System R&D Center, Mitsubishi Electric Corporation, Kamakura, Japan.
Since 2001 he has been engaged in research and development of IP-optical networking and all-optical networking.

P-4 "IGP Extentension for RWA in WSON"
Jianrui Han, Dan Li, Jianhua Gao, Young Lee Huawei Technologies, P.R.China

With the deployment of the Reconfigurable Optical Add-Drop Multiplexer (ROADM) and the Wavelength Selective Switch (WSS), WDM networks have become more dynamic (Figure 1). In WDM network, RWA is used to calculate an available wavelength path with wavelength information of the network. To calculate a wavelength path accurately, the information should contain not only topology and the TE link states but also wavelength connectivity, wavelength conversion capability and per-fiber wavelength availability. To distribute these constraint information in WDM network, one approach is to extend GMPLS IGP-TE. However, due to the size of all these constraint information, a compact encoding of representing these information is important. This presentation introduces a solution to distribute the constraint information with least possible data, especially for the wavelength connectivity, so as to not affect the performance of IGP.

To represent the wavelength connectivity and wavelength conversion capability of a WDM node, a matrix is used to indicate potential wavelength connectivity (Figure 2). The matrix is represented by a set of bit map. Each bit represents the connectivity of a specific wavelength in a fiber cross to another wavelength in another fiber (set 0 means “can’t cross connect”). And the per-fiber wavelength availability can also be indicated by bit map (Figure 4). Each bit in the bit map represents whether the wavelength in fiber is available or not (set 0 means “not available”). When the wavelength is used to establish or tear down an LSP, the bit value is changed accordingly. Before advertising the above information, a sequence of wavelength that an optical fiber can support needs to be advertised (Figure 3). The order of wavelength sequence should be consistent with the order corresponding to wavelength represented by the above bit map; therefore this wavelength sequence can be used to parse above bit map information. In this solution, the wavelength information only need to be advertised once while the potential wavelength connectivity and wavelength status information represented by bit map can be advertised whenever changes would occur. It can reduce the data size effectively.

Biography: Jianhua GAO, a system engineer, in Advanced Technology Department of Wireline Network Business Unit of Huawei Technologies Co., Ltd., a leader in providing next generation telecommunications networks. He joined Huawei in 1999. He received his M.Sc in Control Engineering from Harbin Institute of Technology of China in 1999. His research interests include GMPLS control plane、PCE and network planning tool.

P-5 "Establishing Inter-domain services"
Richard Douville, Helia Pouyllau, Alcatel-lucent Bell-labs, France

The future of the Internet will dwell on inter-domain services. Such services (e.g. Video on Demand, VPN etc.) concern critical applications in terms of Quality of Service (QoS) which can not be guaranteed in a best effort fashion.

Problems like negotiating, provisioning and monitoring end-to-end QoS are central issues for inter-domain service establishment. Meeting end-to-end QoS requirements and monitoring them are non-trivial problems studied and partially solved by many authors. However, few of them designed solutions taking into account the economical relations that exist between domains.
On the economical point of view, some carriers may not agree on cooperating with all others to perform end-to-end services. Furthermore, domains have to be administrated independently to one other; and, critical data like resource capacities have to stay private. The contract approach which consists in setting up SLAs (Service Level Agreements) between domains is not a sufficient solution to figure out economical questions. Moreover, SLAs do not solve questions like how to publish them, how negotiating them with respect to domain privacy and cooperation agreements, how to determine penalties in case of contract violations, etc.
Finally, technical heterogeneity have to be overcome: solutions for end-to-end QoS guarantee have to be applicable for any network solution (e.g. DiffServ, GMPLS, etc.).

Hence, in this presentation, we aim to propose a solution for establishing inter-domain services and that handle the following questions:

• Economical relations between domains: they lead us to consider alliance mechanisms (like in air transport). In addition, we propose a centralized architecture where a third party would resell inter-domain services and thus may be an interface between domains and customers.
• Technical difficulties: considering the technical difficulties to overcome, we recommend the use of a PCE-based architecture on the network layer and the implementation of inter-domain and QoS functionalities on the service layer.
• Adaptability: we suggest to enrich the current inter-AS (G)MPLS-TE technology (RSVP-TE, PCEP and policing procedures) in order to enable automatic provisioning of inter-domain TE services. Our purpose is to make these extensions applicable to different contract negotiation frameworks (e.g. classical bilateral negotiation or service layer based negotiation).
• Algorithms: we introduce some algorithmic solutions for inter-domain service level publication and QoS provisioning.
  

Biography: Richard Douville received the M.Sc. degree in electrical engineering from the University of Versailles, France, in 2000. In 2000, he joined Alcatel Research Labs where he has been involved in IP over optical inter-networking issues, including network and system architectures, traffic engineering, resource dimensioning and performance evaluation. His current research interests include control (GMPLS) and management architectures, and solutions enabling the automation of end-to-end inter-domain/layer service deployments.

P-6 "Shared Backup Mesh Protection in PCE-based WSON networks "
Greg Bernstein, Grotto Networks, USA, Young Lee, Huawei, USA

This paper presents shared backup mesh protection control plane and computation strategies in GMPLS and PCE-based Wavelength Switched Optical Networks (WSON). Shared backup mesh protection allows carriers substantial efficiencies in terms of bandwidth as well as wavelength resources saving compared to 1+1 or 1:1 protection schemes. It would also provide satisfactory protection against a single failure. Figure 1 shows a typical scenario for share backup mesh protection scheme.

Figure 1: Network example with Shared Mesh Protection Scheme

In shared backup path computations one computes both a working path and a backup path for the connection simultaneously and in a way that minimizes the amount of resources reserved for the backup path by sharing these resources among other working paths. When applying shared mesh protection to WSON we have constraints that stem from both the wavelength continuity requirement and the backup sharing condition. In addition we have optimality conditions stemming from both the desire to maximize the sharing of backup resources and minimize the total number of wavelengths in use in the network. Finally, we need to keep track of information on a per wavelength basis.

Current approaches [1,2] have had a goal of reducing the amount of additional information that needs to be distributed via a GMPLS routing protocol. While [3] has shown that too little information in the case of a WSON can lead to higher blocking probabilities and higher restoration overbuild. This previous work has focused on what is known as sequential connection requests, however an important application for PCEs is in concurrent lightpath computation. In this paper we look at the information needed in shared backup path computations in WSONs and suggest a server-based PCE approach that keeps track of the mapping information between working paths and shared backup paths. This approach could aid in reducing the amount of information that would need to be flooded.

[1] M. Kodialam, M. Kodialam, and T. Lakshman, “Dynamic routing of restorable bandwidth-guaranteed tunnels using aggregated network resource usage information,” Networking, IEEE/ACM Transactions on, vol. 11, 2003, pp. 399-410.
[2] Guangzhi Li et al., “Efficient distributed path selection for shared restoration connections,” INFOCOM 2002. Twenty-First Annual Joint Conference of the IEEE Computer and Communications Societies. Proceedings. IEEE, 2002, pp. 140-149 vol.1; http://ieeexplore.ieee.org/iel4/7943/21921/01019255.pdf.
[3] R. Martinez, R. Munoz, and R. Casellas, “Experimental Shared Path Protection Algorithms in Distributed All-Optical GMPLS-based Networks,” Proc. 6th International Workshop on the Design of Reliable Communications Networks (DRCN), 2007.

P-7 "P2MP Path Computation & Operation in MPLS & Optical Networks"
Quintin Zhao, Huawei, USA, Daniel King, Aria Networks, UK, Fabien Verhaeghe, Marben, France, Tomonori Takeda,NTT, Japan

In this presentation, we discuss an application of PCE-based operation of point-to-multipoint (P2MP) label switched paths (LSP) in MPLS and optical networks. Path computation for P2MP traffic-engineered (TE) LSPs presents a significant challenge because of the complexity of the computation. Determining disjoint protection paths for P2MP TE LSPs can add considerably to this complexity, while small modifications to a P2MP tree (such as adding or removing just one leaf) can completely change the optimal path. Reoptimization of a network containing multiple P2MP TE LSPs requires considerable computational resources. All of this means an ingress label switched router (LSR) may not have sufficient processing power to perform the necessary computations, and even if it does, the act of path computation might interfere with the control and management plane operation necessary to maintain existing LSPs. We will summerise the latest IETF work, which facilitates offloading such path computations from LSRs to a dedicated network resource capable of computing constrained P2MP paths using a PCE-based architecture.

Fundamental to the determination of the paths for P2MP label switched paths (LSPs) within a network is the selection of branch points. Not only is this selection constrained by the network topology and available network resources, but it is determined by the objective functions that may be applied to path computation. The selection of branch points within the network is further complicated by the fact that not all LSRs in the network are necessarily capable of performing branching functions. Additionally, network policies may dictate specific branching behavior. Alternatively, administration policies may dictate that replication should be concentrated on specific key replication nodes behaving like IP multicast rendezvous points.

The presenter will highlight the PCE-based requirements and solutions for operation of point-to-multipoint (P2MP) label switched paths (LSP) in MPLS and optical networks. These will include the network architecture used to support path computation for P2MP applications, the requirements in support of the path computation for P2MP application and the protocol extensions to the PCEP (PCE communication Protocol) to satisfy the identified requirements.

Biography: Fabien Verhaeghe is Senior consultant for Marben Products for 10 years. Originally as a software programmer working on OSI protocols and OSI to IP gateways. I am now responsible for the Marben Products GMPLS routing protocols software (OSPF-TE, ISIS-TE and PCE) and am in charge of keeping those products up to date by following the standardization process. I am currently involved in the Point to Multipoint PCEP extensions in the IETF PCE WG.

Tech. Session 2: WSON
Thursday 5, June 2008 16:45-18:25

Chair: Wataru Imajuku, NTT

T2-1 "GMPLS Extensions in Support of All-optical Networking"
Hongxiang Guo, Takehiro Tsuritani, Noboru Yoshikane, Tomohiro Otani, KDDI R&D, Japan

The opaque optical network and its GMPLS-base control plane have been so far claimed to be mature and stable for the deployment. However, due to the ever-increasing bandwidth demands from various emerging applications, advanced and cost-effective all-optical networking technologies are desirable to be gradually introduced. As described in the IETF draft [1], such an all-optical network poses some new requirements on the GMPLS control plane. For example, the high-cost O/E/O regenerators are supposed to be only sparsely placed, and as a result, the wavelength continuity constraint has to be considered in each transparent section. This issue may become even more complicated when a tunable laser is introduced at ingress/egress nodes. Furthermore, it is indispensable to be capable of properly routing a light path through a particular node and explicitly specify the usage of its accommodated regenerator. Otherwise, the light path may fail to be established due to the accumulated physical impairment. Thus, a more effective control plane mechanism incorporating both signaling and routing extensions is preferred, as compared with the previous control plane of the opaque optical network.

Considering these requirements, this presentation investigates the architecture of a typical all-optical network consisting of multi-vendor equipments such as the PXC, ROADM/WXC, transponder and regenerator, and then proposes some potential GMPLS extensions enabling automatic control of all-optical networks. Specifically, a lambda label with global semantic is introduced to simplify the interoperability in this multi-vendor environment [2]. In addition, it is also applicable to insert this wavelength label into ERO in order to explicitly specify the preferred wavelength used at a specific hop such as the transponder at egress node or the intermediate regenerator when the regeneration is needed. This is indicated by defining a new bit-flag named “regeneration bit” in the label sub-object. Furthermore, in conjunction with the definition of such a lambda label, the wavelength mask extended in OSPF-TE can effectively solve the wavelength continuity issue in the path computation process. By using these proposals, the end-to-end light path provisioning and protection by incorporating a tunable laser and the O/E/O regenerator were successfully verified in the all-optical testbed.

In conclusion, we hope the work in this presentation can further promote the standardizing progress of “lambda label” in order to enable a flexible deployment of various all-optical nodes in the existing optical network.

Reference
[1] G. Bernstein, et al., “Framework for GMPLS and PCE control of wavelength switched optical networks”, work in progress: draft-bernstein-ccamp-wavelength-switched-02.txt, Oct 2007.
[2] T. Otani, et al., “Generalized labels of lambda-switching capable label switching routers (LSR)”, work in progress: draft-otani-ccamp-gmpls-lambda-labels-01.txt, Nov 2007.

Biography: Hongxiang Guo is received the B.S. and Ph.D. degrees in electrical and communication engineering from Beijing University of Posts and Telecommunications (BUPT), Beijing, China, in 2000 and 2005, respectively. He is currently with KDDI R&D Laboratories, where his work is focus on the algorithm design and performance analysis in optical burst/packet switching network, and the control and management technologies for photonic networks.

T2-2 "Optical Reach Computation with GMPLS"
Snigdho Bardalai, Richard Colter, Sugiya Hideaki, Fred Gruman, Mike White, Sanjay Gera, Fujitsu, USA

One of the major applications that are being addressed by the Generalized MPLS protocols is the controlling of optical transport networks based on wavelength division multiplexing (i.e. WDM). Today WDM networks are widely being deployed using the reconfigurable optical add-drop multiplexing technology (i.e. ROADM). Deploying such networks involves two basic network planning and design processes - network and link planning and then followed by service planning. Optical service planning is a more complex traffic engineering problem because in addition to the regular constraints, optical signal reach is limited to a certain number of spans and network elements due a number of factors that depend on the physical characteristics of the optical devices and fiber in use. This is due to the impairments of the signal over longer distances.

From an operational perspective one issue has to do with the accurate inventory (i.e. planned versus actual) and state, which may include items such as fiber length, of the network resources that are being used to carry out the optical reach estimations. Distributed approaches can help address some of these issues since it eliminates the synchronization of the network inventory and state with a stand-alone server. The distributed solution that is being presented in this paper is based on GMPLS concepts. Optical information related to optical signal to noise ratio (OSNR), polarization mode dispersion (PMD), residual dispersion (RD) etc. for each node and span is advertised. Using this information and given a specific route (i.e. a sequence of nodes and links) it is possible to estimate the optical signal reachability.

The other aspect that this paper is addressing is the end-to-end operational flow and the application of this function in network management scenarios.

Outline
Optical traffic engineering problem statement
Relation between routing and wavelength assignment (RWA) and optical reach
Call-flow and application model
Optical link data model
Routing protocol advertisements
Optical reach criteria
Future topics - route optimization, regeneration sites, inter-domain solutions

Biography: Snigdho Bardalai is a Distinguished Systems Engineer working for Fujitsu Network Communications Inc. located in Plano, Texas, USA. He has been working for Fujitsu for the last 11 years. He is a subject matter expert in optical control-plane technologies and is the lead architect for the control-plane development. He is an active follower of the IETF CCAMP and PCE working groups. He has also contributed to several CCAMP drafts. He holds a Masters degree in Computer Science and a Bachelors degree in Electronics and Communications Engineering.

T2-3 "GMPLS perspectives to control WSON"
Pierre Peloso, Benoit RONOT, Martin VIGOUREUX, Richard DOUVILLE, Alcatel-Lucent, France

Wavelength Switched Optical Networks (WSON) consists in the transport of optical circuit connections, without O-E-O conversion through photonic nodes. The induced signal continuity brings technological challenges to control plane protocols. Namely, wavelength assignment and physical limits of the transmission reach. Wavelength assignment is the logical setup of a LSP with the same wavelength all along the route, and the placement of re-coloration points. Transmission reach limitations occur in widespread network where long distance LSP shall use O-E-O regeneration in some intermediate nodes to insure its signal quality (error-free). O-E-O regeneration is needed because WSON deployments are not limited to transparency islands.

GMPLS current state contains basic specialisation to support WSON constraints. We believe the extent of these specialisations is too limited today. Let us consider a typical WSON GMPLS implementation based on a two steps process of Routing Wavelength Assignment (RWA):

• OSPF-TE describes network links (LSA), which are used to compute a spatial route.
• The spectral routing (wavelength assignment) is achieved in a distributed manner through RSVP-TE signalling, wavelength continuity being ensured through label swapping mechanism.

In ideal WSON world, GMPLS would embed additional features to better address aforementioned specificities.
A first mechanism would help determining whether signal regeneration is needed or not. This certainly implies carrying some parameters describing the physical properties of links either during routing or signalling. A joint mechanism should also be able to allocate O-E-O regenerators when building LSPs.
As far as wavelength assignment is considered, limitations appear when no wavelength is available all along a route. Currently carried information is insufficient to determine a correct placement for re-coloration. Concurrent setup of PROTECT and WORKING LSPs stresses even more these limitations.
Two options can be considered to solve these issues:

• Option 1: Perform wavelength assignment jointly with route calculation in the ingress node. This solution demands the flooding of a detailed view of the network (wavelength usage, nodes switching capacity). This questions the scalability of the solution, and demands much standardization work. Nevertheless, the high bandwidth of WSON LSPs limits their dynamicity and this solution facilitates optimal resource usage and path computation.
• Option 2: Enrich the signalling phase. The path message containing a Generalized_Label_Request should convey information about re-coloration resources, helping wavelength assignment, which happens after spatial routing computation.

Finally, a helpful complementary feature for WSON GMPLS can be the global semantic given to wavelength label (e.g. channel frequency). This eases both inter-operability and usage of Explicit_Label_Control to synchronize the setup of multiple LSP sharing wavelengths (bidirectional or protection patterns).

This presentation will survey IETF initiatives in the perspective of the aforementioned WSON issues. It would detail the strength and weaknesses of contributions before concluding on the evolutions brought to GMPLS framework.

Biography: Richard Douville received the M.Sc. degree in electrical engineering from the University of Versailles, France, in 2000. In 2000, he joined Alcatel Research Labs where he has been involved in IP over optical inter-networking issues, including network and system architectures, traffic engineering, resource dimensioning and performance evaluation. His current research interests include control (GMPLS) and management architectures, and solutions enabling the automation of end-to-end inter-domain/layer service deployments.

T2-4 "Wavelength Switched Optical Networks with Distributed Control"
Hiroaki Harai and Sugang Xu, National Instituteof Information and Communications Technology, Japan

In a foreseeable future, wavelength paths will be dynamically provided to users on-demand basis in a wavelength switched optical network (WSON). The number of wavelengths multiplexed and the number of nodes in the network will be large. Distributed control mechanisms for wavelength decision and path selection play very important roles in the robustness and fast path setup for the WSON.

We developed a small-scale WSON with distributed control system recently. This consists of optical equipment such as photonic cross-connects (PXCs), arrayed waveguide gratings (AWGs) for wavelength multiplexing/ demultiplexing, optical amplifiers, and optical transceivers each of which wavelength is compliant with 100 GHz interval specified in ITU-T G.694.1. Each host is capable of multiple Gigabit Ethernet interfaces (with different wavelengths) for D-Plane and an Ethernet interface for C-Plane. A distributed control system consisting of a link-state advertisement mechanism, a path calculation mechanism, and a path setup mechanism is implemented. Lambda switch capable (LSC) function is available in part.

We here propose a distributed control mechanism for wavelength path setup. This is done by using existing RSVP-TE for GMPLS and local internal procedures at edge (or user) and core nodes. For reservation, a wavelength is decided at the destination node by a local policy (i.e., a domain-specific rule such as random, first-fit, ranking, etc). More importantly, a set of wavelengths in the Label Set in the Path message are selected at the source node by a local policy. In a case of no policy at the source, all wavelengths may be included in the Label Set object. This is likely situation to avoid wavelength-resource starvation. However, the number of wavelengths in the Label Set should be kept small for alleviating processing burden of intermediate nodes and for avoiding blocking due to the same wavelength reservation by the simultaneous different wavelength-path requests. The number of wavelengths in the Label Set object changes according to some measurement results at the source node. That’s why we call it a distributed system.

For link-state advertisement, wavelength-state information (i.e., busy/idle state for each wavelength on a fiber) is optionally advertised for reflecting signaling result. Link-state change from both unidirectional and bi-directional path setups/releases is supported. For the experimentation purpose, OSPF-TE for GMPLS is further extended. In this talk, we will present our WSON testbed and an RSVP-TE friendly distributed signaling mechanism, which was summarized in this abstract.

Biography: Hiroaki Harai is currently a Project Leader of National Institute of Information and Communications Technology (NICT), Tokyo, Japan, where he is leading Optical Grid Infrastructure Project. His interest also includes design of new generation network architecture and optical circuit and packet switch integrated network. He received the M.E. and Ph.D. degrees in Information and Computer Sciences from Osaka University, Osaka, Japan. He was elected to Outstanding Young Researcher in the 3rd IEEE ComSoc Asia-Pacific Young Researcher Award. He is concurrently a Visiting Associate Professor of The University of Electro-Communications, Tokyo, Japan.