IP over Optical Working Group Internet Draft Expiration Date: April 2004 Xu Shao Institute for Infocomm Research Tee Hiang Cheng Institute for Infocomm Research Nanyang Technological University Kumaran Veerayah Institute for Infocomm Research October 2003 Requirements for MPLS over GMPLS-based Optical Networks (MPLS over GMPLS) draft-xushao-ipo-mplsovergmpls-00.txt Status of this Memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC2026. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet- Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. Abstract MPLS over GMPLS-based optical networks (MPLS over GMPLS) is a subset of IP over optical networks. To be more specific, in this draft it refers to the technology of interconnection between MPLS networks and Xu Shao et al, Expires - April 2004 [Page 1] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 GMPLS-based optical networks with an overlay model or an interdomain model. It is an important milestone in the evolutionary roadmap from IP over static WDM to a peer model of network interconnections. The most significant feature of the requirements for MPLS over GMPLS is a much more dynamic interface between the two layers. The draft discusses the evolutionary roadmap of IP over optical networks and then highlights the significance of the concept of MPLS over GMPLS. Some new requirements will be identified, including multi-lightpath connections, MPLS network topology dynamic changes and dynamic traffic grooming and so on. It is these requirements that bring some challenges to the present routing, signaling and UNI protocols. Table of Contents 1. Summary for Sub-IP Area........................................2 1.1 Summary....................................................3 1.2 Where does it fit in the Picture of the Sub-IP Work........3 1.3 Why is it Targeted at this WG..............................3 1.4 Justification of Work......................................3 2. Specification of Requirements..................................3 3. Introduction...................................................3 3.1 Terminology................................................4 4. Overview of MPLS over GMPLS Service Model and Requirements.....5 4.1 Evolutionary Roadmap of IP over Optical Networks...........5 4.2 Overview of MPLS over GMPLS................................6 4.3 Why MPLS over GMPLS?.......................................7 4.4 Requirements for MPLS over GMPLS...........................8 5. Dynamic Use of Multi-lightpath Connections between Two LSRs....9 6. Dynamic Topology Changes of MPLS Networks.....................11 7. Dynamic Traffic Grooming......................................12 8. Virtual Wavelength Assignment (VWA) Problem...................14 9. MPLS Survivability versus GMPLS Survivability.................14 9.1 MPLS survivability Only...................................15 9.2 GMPLS survivability Only..................................15 9.3 Integrated Survivability..................................15 9.4 QoS Mapping...............................................15 10. Topology Driven Label Assignment in MPLS over GMPLS Networks.15 11. Multicast in MPLS over GMPLS Networks........................16 12. Interdomain Interconnections.................................16 13. Security Considerations......................................16 14. Acknowledgements.............................................16 15. References...................................................16 16. Author's Addresses...........................................17 17. Full Copywrite Statement.....................................18 1. Summary for Sub-IP Area Xu Shao et al, Expires - April 2004 [Page 2] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 1.1 Summary Please see the abstract above. 1.2 Where does it fit in the Picture of the Sub-IP Work This work fits in the IP over Optical (Ipo) working group. 1.3 Why is it Targeted at this WG This draft is targeted at the IPO WG because it specifies the requirements for MPLS over GMPLS-based optical networks, a subset of IP over WDM. MPLS over GMPLS has many new features in requirements that have not been discussed in related drafts so far [IPO-FRAMWORK]. 1.4 Justification of Work The IPO WG should consider this document since it provides many new and practical features in requirements that have not been encompassed by the current requirements of IP over optical networks. 2. Specification of Requirements The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 [RFC2119]. 3. Introduction MPLS over GMPLS-based optical networks (MPLS over GMPLS) is a subset of IP over optical networks. In this draft it refers to the technology of interconnection between MPLS networks and GMPLS-based optical networks with an overlay model or an interdomain model. It is an important stage in the evolution from IP over static WDM to a peer model of network interconnections. Even if someday the peer model of network interconnections is mature, the MPLS over GMPLS model is still very useful and popular for technical and managerial reasons. MPLS over GMPLS has some unique requirements, which are different from the general requirements for IP over optical networks studied in the IP over Optical (Ipo) working group of IETF. MPLS over GMPLS allows a much more dynamic interface between the two layers. In view of the important role played by MPLS over GMPLS now and in the future, it is necessary for us to focus the study on the requirements for MPLS over GMPLS. Xu Shao et al, Expires - April 2004 [Page 3] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 The draft discusses the evolutionary roadmap of IP over WDM and then highlights the significance of the concept of MPLS over GMPLS. Some new requirements will be identified and highlighted in the draft, including multi-lightpath connections, MPLS network topology dynamic changes and dynamic traffic grooming and so on. 3.1 Terminology IP over static WDM (IP over WDM): --------------------------------- In this kind of interconnection model, IP routers are directly connected with lightpaths provided by optical networks. The route computation and wavelength assignment of the lightpaths and the establishment of the lightpaths are performed manually or by a centralized network management system (NMS). The IP networks connected by the optical networks do not support MPLS. MPLS over static WDM (MPLS over WDM): --------------------------------- In this kind of interconnection model, the clients of the optical network are MPLS networks, which have explicit routing and Internet Traffic Engineering (TE) capability. The optical network is still a static network without GMPLS signaling support. IP over GMPLS-based optical network (IP over GMPLS): --------------------------------- In this kind of interconnection model, the clients of the optical network are the traditional IP networks, but the optical network has a proprietary or standard GMPLS-based control plane, which can support dynamic lightpath provisioning and restoration. MPLS over GMPLS-based optical network (MPLS over GMPLS): --------------------------------- In this kind of interconnection model, the clients of the optical network are the MPLS networks. It may be GMPLS-aware or GMPLS- unaware. If an MPLS network can recognize the GMPLS signalings and topology description of optical networks, it is called GMPLS-aware. Otherwise, it is GMPLS-unaware. The optical networks have a proprietary or standard GMPLS-based control plane. There is an UNI between the two networks. Virtual Wavelength Assignment (VWA): --------------------------------- Across the interface between MPLS networks and GMPLS-based optical networks, it may have several wavelength connections. The MPLS network may use part or all the connections at any time according to its instantaneous bandwidth requirements. In order to ensure all the process to be dynamic and avoid any manual operation, the connections Xu Shao et al, Expires - April 2004 [Page 4] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 across the interface must be established in advance, although we need not establish any real lightpaths. In the wavelength convertible GMPLS-based WDM network, we can choose wavelength in random, but in wavelength continuous network, the initial selection of wavelength will affect the optical network performance in the future. We refer this as Virtual Wavelength Assignment (VWA) problem in the context. 4. Overview of MPLS over GMPLS Service Model and Requirements 4.1 Evolutionary Roadmap of IP over Optical Networks The evolvement of IP over optical networks relies on the progress of IP technology, optical network technology and the common control plane technology - GMPLS. The final objective of IP over optical networks is to achieve a dynamic, flexible and resilient network interaction architecture by using a standard common control plane. Thus IP over optical networks has two aspects. One aspect is on the IP-centric common control and measurement plane called GMPLS, and the other aspect is on how to efficiently connect the IP networks with the optical networks. The former aspect is mainly done by CCAMP working group in IETF now. In the draft, we pay more attention to the latter. GMPLS is a universal control plane not only for IP networks but also for WDM optical networks. It can support overlay model, interdomain model and peer model of IP over optical networks. GMPLS may not be achieved in one step. Therefore, it is important to find the roadmap of evolution. The roadmap can be summarized as follows: -- Initially, IP over static WDM (IP over WDM). In this stage, the lightpaths required by IP routers are configured manually or by a centralized network management system. There is no signaling involved in the total process in terms IP network, optical network and their interfaces. This is a static overlay interconnection model for IP over optical networks. -- Next, MPLS over static WDM (MPLS over WDM), or IP over GMPLS-based optical networks (IP over GMPLS). With the independent evolution of IP network technology and optical network technology, the architecture of data communication networks and optical communication networks are changing significantly. The next stage for IP networks is the enhancement of MPLS capability, while he next stage for WDM optical network is a GMPLS-based optical network, which can dynamically provide lightpaths and restoration. During the course of evolvement, IP over WDM becomes either MPLS over WDM or IP over GMPLS. Generally speaking, in this stage, there may be a simple UNI signaling, but sometimes manual operation and configuration is unavoidable. Xu Shao et al, Expires - April 2004 [Page 5] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 -- Then, MPLS over GMPLS based optical networks (MPLS over GMPLS). In this stage, the optical network can support GMPLS and the IP network is upgraded into the MPLS network. Sometimes, the interconnection model between the two layers is an overlay model or a loose inter- domain model. This is a very important milestone, which enables us to set up LSPs (MPLS LSPs or optical LSPs) across the interfaces totally automatically on the basis of optimization of data and optical network resources by taking advantage of the traffic engineering capabilities from MPLS as well as GMPLS. -- Eventually, integrated GMPLS networks. The premise is that both IP networks and optical networks can support standard GMPLS. At this stage, the two networks can be connected freely, with an overlay model, a interdomain model or a peer model. If the overlay model or interdomain model is used, in terms of requirements, actually there are no essential differences from the MPLS over GMPLS in the last stage. That is why to study the requirements for MPLS over GMPLS is so meaningful. The following figure shows the evolutionary roadmap of IP over optical network technologies. IP ----------------------->MPLS ---------------------->GMPLS network network Aware | \ / | | | \ / | | IP over IP over GMPLS MPLS over GMPLS Integrated WDM or MPLS over WDM | Interconnection | / \ | | | / \ | | Static ------------>Partially standard GMPLS------------>GMPLS WDM network |<- Stage I ->|<--Stage II-->|<--Stage III-->|<---Stage IV---->| 4.2 Overview of MPLS over GMPLS As discussed above, Generally MPLS over GMPLS refers to an overlay model or inter-domain model of interaction between GMPLS-based optical network and MPLS-based Internet. For the overlay model, from the perspective of the MPLS networks, MPLS LSRs are connected by the Optical Virtual Private Networks (OVPNs). All the LSRs connected by the optical networks can belong to the same routing area or the same autonomous system (AS). For the interdomain model, the optical network is regarded as an autonomous system, different autonomous systems exchange topology information by some border routing protocols, for instance, BGP4. Xu Shao et al, Expires - April 2004 [Page 6] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 It is necessary for us not only use the traffic engineering and dynamic features in the MPLS network and GMPLS networks internally, but also extend these features across their interfaces. Above of all, once the interfaces are connected, MPLS and GMPLS provide the possibility that these interfaces can be adaptively used according to the bandwidth requirements. Therefore, the connection model between MPLS layer and GMPLS-based optical layer should be a dynamic and adaptive model. The following is an illustration of interconnection of MPLS over GMPLS. There may be multi-lightpath connections between LSR and OXC. Even if there are connections between an LSR and an OXC, it does mean there should be equal lightpaths established in the optical networks, since the LSR may select to use part of all the connections according to its bandwidth requirements. Moreover, an LSR connected with an OXC may try to request a lightpath with any LSRs connected by the optical networks if it has its identifiers, depending the interconnections models. +-------------+ +----------------------------+ +-------------+ | +---+ | | +---+ +---+ +---+--+====+-+---+ | | +---+LSR+-+----+--+OXC+----+OXC+-----+OXC+--+====+-+LSR+---+ | | | +-+-+ | | +---+ /+---+ +---+--+====+-+-+-+ | | | +---+ | | | | / GMPLS-based | | | | +-+-+ | | |LSR| | | | | / WDM Networks | | | | |LSR| | | +-+-+ +-+-+ | | +---+/ +---+ | | +-+-+ +-+-+ | | +---+LSR+-+----+--+OXC+--------------+OXC+--+----+-+LSR+---+ | | +-+-+ | | +-+-+ +---+ | | +---+ | | MPLS | | +-----\----------------/-----+ | MPLS | +---------+---+ \ / +-------------+ | +---------\--------/---------+ | | \+---/ | | | +------+LSR|-----+ | | | | +---+ | | | | +---+ MPLS +---+ | +--------+---+LSR+------------+LSR+ | | +---+ +---+ | +----------------------------+ 4.3 Why MPLS over GMPLS? There are enormous reasons that make MPLS over GMPLS so attractive. -- Even if a peer interconnection is possible, it is not always very competitive due to its complexity in management and insecurity. A functional division is necessary. Sometimes, a well-designed MPLS over GMPLS architecture and interfaces can achieve the same optimal objectives as a peer interconnection model is able to provide. Xu Shao et al, Expires - April 2004 [Page 7] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 -- MPLS network may not recognize the GMPLS protocols, i.e., the MPLS network is GMPLS-unaware. In this scenario, the network architecture is unable to be built with peer model. -- The two networks sometimes belong to two different network operators. The optical network operator does not want the customers to have its topology for security reason. -- In IP over GMPLS, the interface between IP networks and GMPLS networks can only be dynamically used by the router directly connected with the optical networks. Other routers have no means to control the interface, although they may be aware of the interfaces. With the enhancement of traffic cotrol ability in MPLS, the interface will be able to be dynamically used by either LSRs in the MPLS networks. This enables the dynamic use of the MPLS/GMPLS interfaces to optimize the network resource from the perspective of total network, on a node or interface. -- As a connectionless routing network, IP has not provided the measures to use backup or restoration paths. It relies on the survivability from layer 2 and layer 1. Therefore, in an IP over GMPLS networks, survivability should mainly depend on optical layer. Working between the layer between layer 2 and layer 3, MPLS itself can support flexible backup or restoration. Thus, MPLS over GMPLS networks have the flexibility to choose from either or both. It is crucial to study how to combine them together to achieve cost effectiveness and scalability. 4.4 Requirements for MPLS over GMPLS MPLS is the enhancement of IP protocols in traffic engineering, explicit routing and QoS etc. Compared with IP over WDM, IP over GMPLS, the main significant enhancement in MPLS over GMPLS is that it makes it necessary and possible to support the more dynamic interactions between the two layers. The dynamic interactions are able to be controled by either LSRs across the MPLS networks. Sometimes, these new features should be achieved by enhancing respective protocols. In the following sections, we will study these new requirements for MPLS over GMPLS, including -- Dynamic use of multi-lightpath connections between two LSRs -- Dynamic topology changes of MPLS networks -- Dynamic traffic grooming -- Virtual Wavelength Assignment (VWA) problem -- MPLS survivability versus GMPLS survivability -- Topology driven label assignment in MPLS over GMPLS networks -- Multi-cast in MPLS over GMPLS networks Xu Shao et al, Expires - April 2004 [Page 8] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 -- Interdomain interconnections 5. Dynamic Use of Multi-lightpath Connections between Two LSRs In MPLS over GMPLS networks, two LSRs can be connected by more than one lightpaths, i.e., multi-lightpath connections. This allows the LSRs to dynamically decide to use one or more lightpaths according to their bandwidth requirements, a cost-effective way for both optical network operator and its clients. Nowadays, the typical bandwidth of a wavelength is from 2.5 Gbps to 10 Gbps. A fiber in a commercialized system can typically support up to 160 channels. Generally speaking, it is difficult for IP network to use multi-link connections between two routers due to the limitation of IP protocols. Therefore, in IP network, to cope with the traffic growth between two routers, the usual way is to upgrade the interface bandwidth between the two routers. Given a wavelength in an optical WDM network is fixed to 2.5Gbps, in traditional IP networks, generally two adjacent routers can only be directly connected with one wavelength. Once the two routers need a 10Gbps interface, we have to either directly upgrade the wavelength to 10Gbps or redesign the topology of the IP networks by routing some IP packets via other routes. Fortunately, it is easy for MPLS to support parallel links between two LSRs and even balance the traffic among all the links. Similarly, if the two LSRs are connected by lightpaths provided by GMPLS network, MPLS can support multi-lightpath connections. Note that the multi-lightpath connections may have different routes in optical domain. Thus, if the pair of LSR needs a 10Gbps lightpath, alternatively, we can establish 4 parallel lightpaths, each of which is 2.5 Gbps. This is a very significant feature of MPLS because it allows the dynamic usage of the wavelengths according to its bandwidth requirements between the two LSRs. If the two LSRs have more traffic, they can use more lightpaths. Otherwise, they can release some lightpaths to save cost. This is a cost-effective method not only for MPLS network but also for optical networks. With MPLS protocols, either LSR can determine to use any quantities of wavelength according to their requirements. Lightpaths are dynamically set up by GMPLS protocols driven by the arrival of MPLS LSPs, or dynamically torn down driven by the release of MPLS LSPs. The MPLS Label Switching Paths are nested to the lightpaths, constructing an LSP hierarchy. In the MPLS over GMPLS networks, all the procedures are expected to be totally automatic in terms of the establishment of MPLS LSP or optical lightpath. The advantages of multi-lightpath connections can be summarized as: 1. Cost effective. For the optical network operator, throughput will be improved and thus the operator can get more operating revenue from lightpath provisioning service. For the subscribers, costs will Xu Shao et al, Expires - April 2004 [Page 9] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 decline significantly since they do not need to pay a lightpath unless they use it. 2. Survivability. The multi-lightpath connections can be scheduled to use different routes, such as shared risk link group (SRLG)-disjoint. That means from the perspective of subscribers the survivability has been enhanced. Once there is a breakdown of a lightpath, other lightpaths are still available. 3. Easy to access. When a pair of multi-lightpath connection is required, initially we need not really set up all the optical lightpaths, but one or part. This makes the optical network easy to solicit all kinds of customers with different maximum bandwidth requirements. 4. Seamless integration between electrical and optical layer. In traditional network architecture, MPLS and GMPLS are both automatic and intelligent networks in terms of resource usage, but the interface between the two layers is fixed and dumb. Now we extend those automatic and intelligent features to the interface. In summary, it is really convenient that MPLS network can request or release any lightpaths dynamically. But we are facing the risk that part or all the lightpaths may be unavailable when they are required as the optical network is a blocking network. Fortunately, this usually does not bring any significant loss in MPLS network due to the traffic engineering capability provided by MPLS. For example, MPLS network can select other alternative route if it finds that one route is too crowded. Another method is to try to make use of other available lightpaths by changing MPLS network topology, which will be discussed in the next section. But if the LSPs in MPLS networks have QoS requirements, the unavailability of in setting up lightpaths may not ensure the QoS of every LSP due to the limited bandwidth resources. Therefore, some tradeoffs must be made to reserve some lightpaths for future use. If the MPLS networks can predict requests for more bandwidth in the near future, it should try to establish part or all lightpaths in advance. This is especially necessary when the blocking probability in the optical networks becomes higher. Another way maybe use the topology change method discussed in the next section. As a result, it is necessary to enhance the UNI signaling for better support multi-lightpath connections in MPLS over GMPLS networks. 1. It is necessary for the MPLS network to know the states of the optical network, such as blocking probability, by the exchange of UNI signaling. After knowing these states, the MPLS network can positively use corresponding policies to avoid the loss when its Xu Shao et al, Expires - April 2004 [Page 10] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 lightpath requests are blocked. One policy is to reserve some lightpaths in advance prepared for future use. 2. The multi-lightpath connections may have different QoS requirements. It is necessary to map different QoS requirements to the route selection of lightpath. So in the UNI signaling, some parameters should be added to enable the necessary QoS mapping between two networks. 6. Dynamic Topology Changes of MPLS Networks From a traditional IP network point of view, network topology has to be kept unchanged as long as possible. Any topology change will invoke a link state advertisement (LSA) flooding process. It will take considerable time for all the routers in the area to update respective routing tables. As a result, during the unstable state, it may cause some congestion in some nodes or links. Due to the lack of traffic engineering capability, the topology after change may not be well designed to route all the traffic uniformly in the total network. Hence, topology change is generally regarded as a transitory process and unstable network state in traditional IP networks. MPLS is challenging this concepts due to its new features compared with connectionless paradigm in traditional IP networks. For the MPLS network, if we do not consider topology driven, MPLS can support topology change, because explicit routing is used and traffic engineering capability can balance the traffic from the whole network perspective. Therefore, MPLS is insensitive to the topology change. We can positively use topology change to solve the problem discussed in last section. Multi-lightpath connection discussed above is only one way to try to use lightpaths adaptively and cost effectively. Topology dynamic change is another way. If an interface in an LSR needs more bandwidth, it can try to establish lightpath to any counterparts connected on the optical network if possible. We can change the MPLS not only by dynamically decreasing or increasing its interface bandwidth but also by dynamically simplifying or expanding its topology. The method has many advantages: 1. If an LSR has not any lightpath resources to the intended LSR, it can try to connect other LSRs connected with the optical network. This will lower the possibility of blocking. 2. Cost effective. For the optical network operator, throughput will be improved due to lower blocking probability and thus the operator can get more operating revenue from lightpath provisioning service. For the subscribers, costs will decline significantly because they do not need to pay a lightpath unless they use it. Xu Shao et al, Expires - April 2004 [Page 11] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 3. Survivability. Once there are some faults on one UNI, the LSR may not keep connected with the optical networks any longer. In this scenario, its counterparts can establish lightpaths with other LSRs on the optical networks. The dynamical adjustment of MPLS network topology will minimize the effect by node or link failures. The UNI failures of a pair of LSR connected by an optical network only affects one LSR, other than two. The problem of topology change is that it takes some time to converge by LSA flooding. So we do not expect very frequent topology change unless very necessary. This method does not suitable for topology driven which will be discussed in section 10. 7. Dynamic Traffic Grooming SONET/SDH can content with different granularities of bandwidth requirements. However, WDM optical network only provides lightpaths with fixed bandwidth. Nowadays, few subscribers are willing to employ such high bandwidth, so traffic grooming is an efficient and cost effective way to maximize the revenue of one wavelength by combining more low-speed requests into one lightpath. Traffic grooming refers to techniques used to pack low-speed traffic streams onto high-speed wavelengths in order to minimize the network wide cost in terms of line terminating equipment and/or electronic switching. In MPLS over GMPLS network, the most significant objective for traffic grooming is no longer the minimization of the amount of electronic devices, but the maximization of optical network operating revenue and minimization of the cost for using the lightpaths from the MPLS network perspective. MPLS over GMPLS does not involve any hardware costs and all the process expects to be completed automatically without any manual operation, which makes traffic grooming MPLS over GMPLS very interesting. Traffic grooming in ring SONET/WDM networks and mesh WDM networks has been studied intensively in recent years. However, as a dynamic and automatic network, there are many new features in traffic grooming in MPLS over GMPLS networks as follows: 1. In GMPLS, the required lightpath is dynamically established by signaling protocols. Grooming should be performed at the edge, so a lightpath is unable to be terminated to added some requests even if the load on the lightpath may be very light. Thus, grooming in GMPLS is single-hop grooming. The GMPLS does not support multi-hop grooming and it is difficult to extend GMPLS protocols to support multi-hop grooming. This is the main difference from traffic grooming in static WDM mesh networks. Therefore, numerous studies on multi-hop grooming and virtual topology design in mesh WDM networks cannot be used in GMPLS networks. This feature makes the capability for Xu Shao et al, Expires - April 2004 [Page 12] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 grooming in GMPLS very limited compared with multi-hop grooming in mesh WDM networks. 2. The opportunity to groom low-speed requests in GMPLS network is much lower than in ring networks. For a mesh network, the new requests may destine at any nodes in the network. It may not have any pre-configured lightpaths along the path at that time when a request arrives, depending on the load of the network and the scale of the network. 3. Traffic grooming in MPLS over GMPLS is dynamic grooming, which means not only the connection requests arrive randomly, but also the groomed traffic in a wavelength may terminate at any time. However, as long as there are connections existing, the lightpath will have to be held. As a result, dynamic grooming makes grooming complex and not always a cost effective way to make full use of network resource. GMPLS is a dynamic network always with certain blocking probability. Sometimes we will face the risk that when a big customer arrives, there are not enough resources to set up a lightpath because the key resources are occupied by some small customers. So we are facing this embarrassing situation: On one hand, we wish to accommodate more subscribers no mater how much bandwidth they require; on the other hand, we need to give preferential treatment to big subscribers to maximize revenue, especially in the condition of limited network resource. In fact, this is a fundamental conflict between short-term network operating revenue and long-term network operating revenue. To ensure that the long-term network revenue is as high as possible, some low bandwidth requests and low bandwidth requests for grooming should be rejected to make room for future high bandwidth requests. In MPLS over GMPLS network, traffic grooming will be accomplished at the UNI and all the process is expected to be dynamic. In an integrated interaction model, a lightpath will be announced as a forwarding adjacency (FA). Traffic grooming is achieved by establishing a hierarchy of LSPs. In an overlay model, the lightpath is regarded as a point-to-point link. To support dynamic traffic grooming in MPLS over GMPLS networks, there are some requirements on the UNI. -- All the LSPs in a lightpath may not have the same recovery requirements, so it is necessary to have some measures in the UNI to bundle the same kind of LSPs together, and then map these requirements in MPLS layer onto optical layer. -- All the LSRs in an MPLS domain should have the information on how much bandwidth a lightpath remains so as to decide if it is possible to groom a current LSP request. Xu Shao et al, Expires - April 2004 [Page 13] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 -- A fragment bandwidth management policy is very important. Sometimes it is necessary to reserve some bandwidth for the forthcoming high bandwidth LSPs. -- To prevent long time occupation by small bandwidth LSPs, it is necessary to have a policy to manage these small bandwidth and long holding time LSPs. -- There should be some tradeoffs and optimization between setting up a new lightpath or grooming it into a present lightpath, when an LSP request comes. 8. Virtual Wavelength Assignment (VWA) Problem It is essential to design the interface between client layer and optical layer to make the blocking probability as low as possible. In the case of multi-lightpath connections, if the pair of LSR needs up to 4 lightpaths, we must configure 4 wavelengths in advance to avoid any manual operation in the future, even if they may not need the real establishment of corresponding lightpaths. We do not expect to change the traditional interface architecture between client layer and optical layer, in which clients are directly connected with wavelengths without complex switching fabrics. Hence, which wavelengths to be selected must be determined in advance, and thereafter, all the requested wavelengths will have to select wavelengths from these pre-configured wavelengths. How to perform the wavelength assign at this stage will affect the network performance in their future. We refer this as Virtual Wavelength Assignment (VWA) problem in the draft. In the wavelength convertible GMPLS-based WDM network, we can choose wavelength in random. But in wavelength continuous network, the future wavelength assign will have to be confined within the initial configuration. This makes the lightpath requests from multi-lightpath connections have higher blocking probability than lightpath for the total optical networks. Therefore, in this case, we must study efficient virtual wavelength assignment methods. To lower the blocking probability in VWA, the preliminary consideration is to ensure the wavelength usage throughout the total optical network distributes as evenly as possible. 9. MPLS Survivability versus GMPLS Survivability As mentioned above, both MPLS and GMPLS can provide network survivability, including protect and restoration. Therefore, there are 3 preliminary approaches to provide network survivability, namely MPLS survivability only, GMPLS survivability only, and integrated survivability from MPLS/GMPLS. Xu Shao et al, Expires - April 2004 [Page 14] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 9.1 MPLS survivability Only With this model, MPLS networks treat GMPLS networks as abstract links. There is no requirements to map the survivability parameters from MPLS layer to GMPLS layer. It may take considerable long time for MPLS layer to detect the fault from from optical layer. 9.2 GMPLS survivability Only With this model, MPLS takes advantage of protect or restoration from GMPLS to overcome the failures from optical layer. MPLS layer protection or restoration always assumes connect ability from optical layer. MPLS totally relays on the GMPLS network to provide survivability across the segment of optical networks. This is in fact an overlay survivability architecture between two layers. This requires the UNI has the ability to allow optical network clients to initialize protect or restoration lightpath. 9.3 Integrated Survivability The integrated model tries to combine the survivability from the two layers together. The couple should be cost effective, flexible and scalable. This requires much more interchange messages between the two networks. For example, if a node from MPLS network knows the topology of GMPLS network, it may use a abstract shortest SRLG protected path. The abstract shortest protected path is an comprehension of hops in both MPLS and GMPLS networks. Nevertheless, it does not mean the physical hops, but an abstract hop taking the costs into account. 9.4 QoS Mapping QoS mapping is to map QoS requirements of MPLS LSPs onto respective lightpaths for the computation of primary lightpath, protect or restoration. This will be even more complicated if a lightpath contains groomed MPLS LSPs that have different QoS requirements. 10. Topology Driven Label Assignment in MPLS over GMPLS Networks In MPLS network, for scalability, some LSPs may be set up or torn down by topology changes. These kinds of LSPs are sensitive to the topology change. Normally, they have not clear bandwidth and QoS requirements on an LSP, so for the requirements of topology driven in MPLS over GMPlS network, they are basically the same as the generic requirements discussed in the IP over optical networks [IPO-FRAMWORK] and the new features discussed in section 7 & 8 are not applicable. Xu Shao et al, Expires - April 2004 [Page 15] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 Sometimes, in MPLS, topology driven label assignment and request driven label assignment are mixed together. In this scenario, the requirements discussed above still can be used in the LSPs from request driven label assignment. 11. Multicast in MPLS over GMPLS Networks Under consideration. 12. Interdomain Interconnections Generally speaking, the new features discussed above are suitable for not only overlay models, but also interdomain models. 13. Security Considerations No additional security considerations are beyond the present drafts of IP over optical networks [IPO-FRAMWORK] and OVPN. 14. Acknowledgements We would like to thank Dr. Ye Yabin (I2R), Dr. Cheng Xiaofei (I2R), Dr. Chin Soon Hwa (NTU), Dr. Chai Teck Yoong (I2R), Dr. Zhou Luying (I2R) and Dr. Liu Qiang(I2R) for the discussions with them within the optical network design lab in I2R, Singapore. 15. References 14.1 Normative References [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [MPLS-ARCHITECTURE] Rosen, E., Viswanathan, A., and R. Callon, "Multiprotocol Label Switching Architecture", RFC 3031, January 2001. [GMPLS-OVERLAY] G.Swallow et al., "GMPLS RSVP Support for the Overlay Model," Work in Progress, draft-ietf-ccamp- gmpls-overlay-01.txt. [HIERARCHY] K.Kompella and Y.Rekhter, "LSP Hierarchy with Generalized MPLS TE," Work in Progress, draft-ietf- mpls-lsp-hierarchy-08.txt. Xu Shao et al, Expires - April 2004 [Page 16] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 [IPO-REQS] Y.Xue (Editor) et al., "Optical Network Service Requirements," Work in progress, draft-ietf-ipo- carrier-requirements-05.txt. [IPO-FRAMWORK] Bala Rajagopalan et al., 揑P over Optical Networks: A Framework ?Work in Progress, draft-ietf-ipo-framework-05.txt. [GMPLS-OSPF] K. Kompella and Y. Rekhter, "OSPF Extensions in Support of Generalized MPLS", Work in Progress, draft-ietf-ccamp-ospf-gmpls-extensions-09.txt. [RFC3386] W.Lai, D.McDysan, et al., "Network Hierarchy and Multi-layer Survivability," IETF RFC 3386, November 2002. [IPO-ASON] Aboul-Magd (Editor) et al., "Automatic Switched Optical Network (ASON) Architecture and Its Related Protocols," Work in progress, draft-ietf-ipo-ason- 02.txt, March 2002. 14.2 Informative References [OIF-UNI] The Optical Internetworking Forum, "User Network Interface (UNI) 1.0 Signaling Specification - Implementation Agreement OIF-UNI-01.0," October 2001. [ITUT-G709] ITU-T, "Interface for the Optical Transport Network (OTN)," G.709 Recommendation (and Amendment 1), February 2001. 16. Author's Addresses Xu Shao Institute for Infocomm Research Blk 2, 18 Nanyang Drive Unit 230, Innovation Centre Singapore 637723 Tel: +65 6792 2824 Email: shaoxu@i2r.a-star.edu.sg Tee Hiang Cheng Institute for Infocomm Research Xu Shao et al, Expires - April 2004 [Page 17] draft-xushao-ipo-mplsovergmpls-00.txt October 2003 Blk 2, 18 Nanyang Drive Unit 230, Innovation Centre Singapore 637723 Nanyang Technological University Nanyang Ave, Singapore, 639798 Email: ETHCHENG@ntu.edu.sg Kumaran Veerayah Institute for Infocomm Research Blk 2, 18 Nanyang Drive Unit 230, Innovation Centre Singapore 637723 Email: Veerayah@i2r.a-star.edu.sg 17. Full Copywrite Statement Copyright (C) The Internet Society (2003). All Rights Reserved. This document and translations of it may be copied and furnished to others, and derivative works that comment on or otherwise explain it or assist in its implementation may be prepared, copied, published and distributed, in whole or in part, without restriction of any kind, provided that the above copyright notice and this paragraph are included on all such copies and derivative works. However, this document itself may not be modified in any way, such as by removing the copyright notice or references to the Internet Society or other Internet organizations, except as needed for the purpose of developing Internet standards in which case the procedures for copyrights defined in the Internet Standards process must be followed, or as required to translate it into languages other than English. The limited permissions granted above are perpetual and will not be revoked by the Internet Society or its successors or assigns. This document and the information contained herein is provided on an "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Xu Shao et al, Expires - April 2004 [Page 18]