Service Adaptation and Label Forwarding Mechanism for MPLS-TP

This work was funded by the National High Technology Research and Development Program of China (“863” Program) under Grant No. 2008AA01A328.

    Service networks are undergoing transformation due to the shrinkage of traditional telecom services and the trend towards IP-based services. The future market will require packet transport technology that transmits packet services effectively, but also provides carrier-class Operation, Administration and Maintenance (OAM), as well as protection. Transport devices must change from being adaptable to multiservice interfaces into being adaptable to multiservice kernel. In other words, IP-based services require packet transport networks; but at the same time, traditional services such as voice should be delivered normally. To achieve this, packet technologies and transport technologies must be integrated. And in this context, Packet Transport Network (PTN) is introduced[1].

    There are 2 solutions available for PTN implementation: enhanced Ethernet technologies, and transport technology combined with Multiprotocol Label Switching (MPLS). Provider Backbone Bridge Traffic Engineering (PBB-TE)—or Provider Backbone Transport (PBT)—is representative of the former; MPLS Transport Profile (MPLS-TP)—or Transport MPLS (T-MPLS)—is representative of the latter. Compared with other technologies, these better serve the goals of PTN, and have come into mainstream use. Both are connection-oriented, have performance and reliability similar to Synchronous Digital Hierarchy (SDH), and offer standard, connection-oriented tunnels. Their main differences lie in the way they implement data forwarding, protection and OAM. In the process of standardization, T-MPLS has outpaced PBT. T-MPLS standards have matured, and meet requirements for commercialization, although there is still room for improvement.

1 MPLS-TP Features and Standardization

1.1 MPLS-TP Standardization
Since 2005, Study Group 15 of the International Telecommunication Union Telecommunication Standardization Sector (ITU-T SG15) has focused on using MPLS technology to define packet transport layer’s service function structure (T-MPLS technology). In 2007, ITU-T made recommendations on aspects such as system framework, interface and device specifications, OAM, protection switching mechanism, and service signal adaptation. These recommendations have received support from the Internet Engineering Task Force (IETF). At present, ITU-T is working on detailed definitions of all layer functions, addition of adaptive client signals, and service interconnection and synchronization. IETF is preparing T-MPLS Request for Comments (RFC), and defining new labels for T-MPLS services. In 2008, ITU-T and IETF established a Joint Work Team (JWT) to promote the merging of T-MPLS and MPLS. IETF expanded existing MPLS technology into MPLS-TP to better support transport requirements specified by ITU-T. In the future, JWT will develop MPLS-TP standards and ensure consistency between T-MPLS and MPLS-TP standards[2].

1.2 MPLS-TP Technical Features
MPLS-TP has a clear framework and complete solutions for implementing its key techniques. It simplifies and innovates on MPLS/Pseudo Wire (PW) technology, introducing concepts such as PTN layer, OAM and linear protection. Thus, it meets the demands of PTN[3]. As a connection-oriented packet transport technology, MPLS-TP consists of a data plane, management plane, and control plane. An end-to-end connection-oriented packet transport tunnel is also established. This tunnel, set up with the network management system or intelligent control plane, is characterized by good operability, and maintenance and protection recovery. MPLS-TP maps client signals into MPLS frames and forwards them using mechanisms such as label switching or label stack. Basic functions of the transport layer—including connection and performance monitoring, survivability and management and control planes—are included[4-5].

    MPLS-TP adopts a dual-label transport mode[6]. When it provides packet data transport for the client layer, it assigns 2 labels to client data: Common Interworking Indicators (CII) and Transport Label Switched Path (T-LSP). The CII label associates the clients at both ends, and is used by terminals to distinguish client data, while T-LSP label is used for exchange and forwarding client data in the MPLS-TP packet data tunnel.

    In short, MPLS-TP is a subset and abbreviated version of MPLS. To support a connection-oriented end-to-end OAM model, many connectionless features are excluded but protection switching and OAM functions recommended by the ITU-T are present. All these facilitate carrier-class services. MPLS-TP also retains MPLS features that are helpful for data service transport, discards the complicated MPLS control protocol cluster defined by IETF, simplifies the data plane, and removes unnecessary forwarding processing. As a result, it is suited to situations where TDM-based services are being transitioned into IP-based ones.

2 MPLS-TP Service Adaptation Technologies
The unified transport/aggregation platform is a breakthrough in service networking technologies. Service providers need not invest in each network layer to transport mixed services. In MPLS-TP, the transport plane is responsible for packet transport of client data, and also adapting to and forwarding client signals. For different client signals, different signal adaptation and forwarding methods are employed. packet data, cell data and TDM data are quite different in length, format, and multiplexing mode. Therefore, different aggregation, fragmentation, encapsulation, sequencing, timing and multiplexing/demultiplexing methods are used in their adaptive transport[7].

2.1 Signal Adaptation
Client signals can be directly mapped into T-LSP, or indirectly mapped based on CII. In the dual-label architecture, all types of service signals can be encapsulated in a dual-label structure. The encapsulation layer structures the specified payload signals for transmission over the Virtual Circuit (VC). It consists of 3 sub-layers: payload convergence, timing, and sequencing. The payload convergence sub-layer is tailored to the specific payload type. It can group a number of payload types into a generic class, and then provide a single convergence sub-layer type that is common to the group. The timing and sequencing sub-layers provide generic services to the payload convergence sub-layer. The functions of the 3 sub-layers are described below.

    (1)  The payload convergence sub-layer is primarily responsible for encapsulating the payload in VC Protocol Data Units (VC-PDUs). It carries the additional information needed to replay the native data units at the Customer Edge (CE) boundary. When a bitstream is sent to MPLS-TP, one part is peeled off the native service processing module. For example, in structured SDH, fragmentation and circuit overheads may be peeled off.

    (2) The sequencing sub-layer performs three functions: frame ordering, frame duplication detection, and frame loss detection. Some services must be delivered in order, while other services need not. The choice of strategy for handling incorrect frame sequence, or detected frame duplication or loss depends on the service type. Client signals such as SDH and Frame Relay (FR) may have  clock recovery and timed packet delivery timing requirements.

    (3) The timing sub-layer performs two functions: clock recovery and timed delivery. Clock recovery involves extracting clock information from the delivered bit stream, and recovering the clock with a phase-locked mechanism. Timed delivery involves delivering non-contiguous VC-PDUs to CEs with a constant phase relation.

    IP services can be directly mapped into T-LSP or indirectly mapped by means of dual labels. In dual-label encapsulation, a node does not need L3 forwarding capability. Hence, in networks where nodes do not have L3 forwarding capability, dual-label encapsulation is advantageous. With non-IP service adaptation, VC-based indirect mapping is adopted. As multiple VC multiplexing is transferred in a Label Switched Path (LSP), the service granularity can be less than 2 Mbit/s. Moreover, with CII and label stack added into the services, address space limitation—which occurs in some technologies—can be overcome.

2.2 Service Encapsulation
The generic encapsulation format of services is illustrated in Figure 1. The payload containing L2 header or L1 overhead can be IP packet, Ethernet packet, ATM cell, FR cell, or SDH payload. Data information and control words are used for payload convergence. With CII label built in, the VC type in T-LSP can be determined; with T-LSP label built in, MPLS-TP LSP can be determined. Control words often include flags, fragment, length, and sequence number. At the destination end—after terminating the LSP and withdrawing the outer T-LSP label—the device decides which high-layer service instance the data stream belongs to. This is based on the inner CII label.

    The transport plane of MPLS-TP has an additional feature: It is transparent to the client layer and service layer. Transparency to the client layer means that any client signal can be carried on MPLS-TP network and transmitted in the form of packets. The client network can be IP, Ethernet, ATM, FR, PDH, or SDH. Transparency to the service layer means that MPLS-TP can use any underlying technology for transport. Before MPLS-TP has its own data link layer protocol, packets can be carried and transmitted via existing Ethernet or SDH networks.

3 Label Forwarding Mechanism
MPLS-TP adopts the dual-label transport mode in which L3 IP is used to transmit L2 Ethernet data. The 2 labels are CII and T-LSP. To support MPLS-TP networking, T-LSP labels can be infinitely embedded. Therefore, there may be several T-LSP labels. The CII label can be a client signal label; for example, in Figure 2, the CII label refers to VC labels. The
multiplexing/demultiplexing module binds several VCs into a VC Group (VCG) and transfers it on the same T-LSP. In this way, the complexity of the transport and switching devices of the network is reduced and less bandwidth is consumed.

3.1 Packet Switching and Forwarding
The service packet switching and forwarding functions of MPLS-TP are mainly involved in exchanging MPLS-TP labels of the packet client data and forwarding the data. Specifically, the functional modules include pre-processing, forwarding and switching, encapsulation, fragmentation, sequencing, timing, multiplexing/demultiplexing, and monitoring.

    (1) Pre-processing refers to the operations on client data before switching. These operations include data and address conversion, and identification of client data type. Pre-processing simplifies the design of the processing steps that follow.

    (2) The convergence module is mainly responsible for sorting and converging packets based on type, and importance of client data signals or signaling signals. It dispatches packets to different transport channels, so that different signal types receive different classes of Quality of Service (QoS).

    (3) The encapsulation module adapts the signals before they are multiplexed with T-LSP and forwarded. The workings of the module are closely related to the client signal type to be encapsulated. For example, different encapsulation methods are used for packets, cells, and Time Division (TD) signals. Encapsulation is the process of attaching VC and T-LSP labels onto the packets, and inserting OAM information into the packets. Depending on the client signal type, 3 sub-modules may be used: fragmentation, sequencing, and timing. If a client signal exceeds the maximum packet length that can be carried by a service layer network, it must be fragmented. Some clients require sequenced delivery and real-time support, so their signals should be delivered in a sequenced and timed way. A sequencing function provides three services: frame ordering, frame duplication detection, and frame loss detection. SDH and FR client signals may require timed clock recovery and timed packet delivery.

    In an MPLS-TP network, the mapping from client signals onto link frames involves client service encapsulation, signal multiplexing, and mapping of MPLS-TP packets into link frames. Figure 3 illustrates the relationship between information units in an MPLS-TP network.

    Client signals can also include OAM packets of T-MPLS networks. Both the data packet and the OAM packets may have a label header for multiplexing. Finally, MPLS-TP packets are mapped onto data link frames, which are transmitted via MPLS-TP topological links. The forwarding and switching module on MPLS-TP terminal devices is responsible for switching the processed client data to related T-LSP and forwarding them. Intermediate transport switching devices continue to forward the packet data with MPLS-TP labels until the data is demultiplexed by the terminal device at the destination. The destination terminal devices forward the data to destination CEs[8-9].

3.2 Signaling Transfer with Dual-Label Transport Mode
To explain signaling transfer in dual-label transport mode, an example will be provided. Suppose there are 2 Ethernet Virtual Local Area Networks (VLANs): Ethernet VLAN 1 (port 1A, VLAN 100) and Ethernet VLAN 2 (port 1B, VLAN 200). An MPLS-TP VC is established between them based on Label Distribution Protocol (LDP). Figure 4 illustrates a solution for signaling transfer between the 2 Ethernet VLANs.

    The process of establishing MPLS-TP VC and transferring signals is described below.
Ethernet VLAN 1 makes a request to Terminal Device 1 of MPLS-TP network to establish an MPLS-TP VC to Ethernet VLAN 2.

    Terminal Devices 1 and 2 of MPLS-TP network negotiate and assign a VC Identifier (VCID) to the VC.

    Terminal Device 1 initializes a LDP signaling session to Terminal Device 2 if one is not already present. Once the two terminal devices receive a LDP KEEPALIVE message from each other, they set up the session and are ready to switch the labels bound on the VC.

    After the Ethernet VLAN’s state becomes “UP”, Terminal Device 1 assigns a local CII label (500 in this example) for the VC based on its VCID. It sets up T-LSP 1 for transport over the VC, assigning the VC a T-LSP label (600 in this example).

 
    Terminal Device 1 converts the T-LSP label into T-LSP tunnel Type-Length-Value (TLV).  the local CII label into LABEL TLV, and the CII Identifier (CII-ID) into Forwarding Equivalence Class (FEC) TLV. It then sends a LABEL MAPPING message to Terminal Device 2.
Terminal Device 1 receives a LABEL MAPPING message from Terminal Device 2 and decodes the message to obtain CII label and CII-ID.

    At the same time, Terminal Device 2 runs the same steps (1) to (6) independently.
After the two terminal devices have finished label binding, exchanging their labels, and confirming their port parameters are correct, T-MPLS VC with CII-ID 50 is successfully set up. If one Ethernet LAN connection fails to work or is removed, a label cancel message will be sent to the opposite terminal device to cancel the assigned MPLS-TP CII label.

3.3 Dual-Label Transport Mode-based VPN Services
MPLS-TP networks have rich OAM overhead functions[10] to monitor and manage signals in the network. This improves the operability and security of the entire network, and paves the way for Virtual Private Network (VPN) services in the MPLS-TP network. Traditional VPN technologies (L2 or L3 tunneling technologies) are easy to implement, but are defective in scalability, security, management and maintenance, QoS, and traffic engineering. In particular, when a client adopts multiple access technologies such as Point-to-Point Protocol (PPP), ATM, FR, and Ethernet, the operator must configure several core networks accordingly in order to offer VPN services. VPN services based on MPLS-TP technology can be implemented in a unified network platform that is independent of transport technologies.

    MPLS-TP VPN is more easily achieved with point-to-point tunnels set up in the MPLS-TP backbone network. Its CEs do not participate in L3 routing, and the client itself configures routing within the VPN. This makes MPLS-TP VPN independent of the client’s L3 network protocol. In MPLS-TP VPN, terminal devices are only responsible for L2 connection, and forwarding between CEs. Functions of L3 or above are performed by CEs. The MPLS-TP network itself can carry different client signals. By means of signal adaptation and encapsulation, a client can use different technologies to access the MPLS-TP VPN, thus satisfying the client’s demand for low cost interconnection of VPNs of different layers.

    Figure 5 illustrates MPLS-TP VPN. It can been seen that dual labels are adopted, where a T-LSP label identifies the shared tunnel between terminal devices, and a CII label identifies the dedicated connection between CEs. The CII label, in the form of a MPLS-TP label stack, is multiplexed in the LSP-based tunnel of the MPLS-TP backbone network. The LSP can be regarded as a tunnel carrying several VCs, and a VC is equivalent to a point-to-point route established by the LDP for VPN users.

4 Conclusion
With good service adaptation and label forwarding capabilities, MPLS-TP technology is well-suited for packet transfer. However, it is rivaled by other packet transport technologies such as PBT. As an emerging technology, MPLS-TP needs the joint efforts of standardization organizations, telecom operators, and equipment vendors to hasten its maturity.

References
[1] 朱京, 刘昭伟, 雷学义, 等. PTN-信息通信基础承载网络的演进与变革[J]. 电力系统通信, 2009, 30(4): 52-59.
      ZHU Jing, LIU Zhaowei, LEI Xueyi, et al. PTN-evolution of information and communication transport network [J]. Telecommunications for Electric Power System, 2009, 30(4): 52-59.
[2] NIVEN-JENKINS B, BRUNGARD D, BETTS M, et al. MPLS-TP requirements [R]. draft-jenkins-mpls-tp-requirements-01. 2008.
[3] ITU-T G.8114. Operation and maintenance mechanisms for T-MPLS layer networks [S]. 2007.
[4] ITU-T Rec G.8131.1. Transport MPLS (T-MPLS) layer network protection switching [S]. 2007.
[5] ITU-T Rec G.8112. Interfaces for the Transport MPLS(T-MPLS) hierarchy [S]. 2006.
[6] ITU-T Rec G.8110.1/Y.1370.1. Architecture of Transport MPLS (T-MPLS) layer network [S]. 2006.
[7] ZHANG Y J, ZHANG Z H, GU W Y, et al. A collision-aware queue assignment scheme for multiservice [C]//Proceedings of the IEEE International Conference on Advanced Infocomm Technology (ICAIT’09), Jul 7-9, 2009, Xi’an, China. Piscataway, NJ, USA: IEEE, 2009.
[8] ITU-T New Supplement Y.Sup4. Transport requirements for T-MPLS OAM and considerations for the application of IETF MPLS technology [S]. 2008.
[9] ITU-T G.8121. Characteristics of transport MPLS equipment functional blocks [S]. 2007.
[10] WARD D, BETTS M. MPLS architectural considerations for a transport profile [R]. ITU-T IETF Joint Working Team. 2008.

[Abstract] Most services borne by transport networks have been transitioned away from Time Division Multiplexing (TDM) towards IP-based. New technology is therefore required to effectively transport packet services. Multiprotocol Label Switching Transport Profile (MPLS-TP) is such a technology which is connection-oriented, capable of multiservice adaptation, and has a flexible label forwarding mechanism. This paper analyzes the characteristics of MPLS-TP, and introduces its latest standardization development. Key issues including data forwarding plane, service adaptation, and label forwarding mechanism are discussed, as well as service implementation of Virtual Private Network (VPN) based on dual-label transport. MPLS-TP applications are summarized in conclusion.