Key Technologies of GMPLS (1)

1 Introduction
With the rapid development of Internet and optical fiber technology, the convergence of IP and optical network technology will inevitably become an important trend of network development in the future. GMPLS provides a good idea for how to solve the fusion between the IP layer and the optical layer. GMPLS inherits almost all MPLS features and protocols and is an extension of MPLS to optical networks. It can use a unified control plane to manage networks created by different technologies, this provides an important guarantee for simplifying the network structure, reducing network management costs, and optimizing network performance.
2 Multi-Protocol Label Switching MPLS) Technology Review 
MPLS is the basis of GMPLS. It is a network technology located at Layer 2.5. It provides a unified operating platform for the interaction between the IP layer and the link layer. It has strong adaptability and flexibility, supports various protocols at the current network layer and link layer, such as IPv4, IPv6, IPX, and AppleTalk for the network layer, and FR, ATM, and PPP for the link layer ). MPLS is a technology that can greatly improve the speed of route forwarding. Its architecture is divided into two independent components, namely, the forwarding component is also called the data layer) and the control component is also called the control layer ). The forwarding component uses the tag forwarding database maintained by the tag switch to perform the forwarding tasks of the Data Group based on the tags carried by the Group. The control component establishes and maintains tag forwarding information between a group of interconnected switches.
The simple working principle of MPLS is: when the data group arrives at the entry LSR Label Switch of the MPLS network Cloud, the entry LSR determines the FEC Forwarding Equivalence Class of the Group by analyzing the information headers of the Data Group, that is, FEC makes some data streams with some commonalities, which are processed by LSR in the same way during the forwarding process), and then searches for the LIB Tag Information Library ), add a label associated with the FEC to the Data Group. In the subsequent LSR, you do not need to search for the IP Group header. You only need to search for LIB Based on the tags of the Data Group to determine its forwarding egress, replace the old tag with the new one before forwarding, and then forward it to the next LSR. When the data group reaches the egress LSR, the egress LSR removes the Label from the Data Group and forwards the data group according to the traditional IP Forwarding method. Among them, all Label Distribution and LSP establishment bound to FEC are completed by LDP Label Distribution Protocol.
3 key technologies of GMPLS

To meet the requirements of smart optical networks to dynamically provide network resources and transmit signaling in the future, we need to expand and update traditional MPLS networks. GMPLS is the product of MPLS expansion to optical networks. It supports traditional group switching, time division switching, wavelength switching, and optical fiber switching, the original routing and signaling protocols are modified and expanded.
At present, the fusion between the IP layer and the Optical Transport Layer mainly involves overlapping models and integration models. GMPLS should support both models at the same time.
The overlapping model is also known as the customer-server model, that is, the optical layer network acts as the server, and the IP network layer acts as the customer layer, which has an independent control plane. Specifically, one is in the core optical network, and the other is in the customer layer, which is embodied in the user-Network Interface (UNI). Route information is not exchanged between the two, and routing is selected independently, it has an independent topology. As a server, the core optical network provides wavelength services for network edge customers. Its advantage is that optical networks and IP networks can develop independently. Its disadvantage is poor network scalability and N2 problems. In addition, two layers have two sets of different address spaces, which requires complex address resolution.
An integrated model is also called a peer-to-peer model or a hybrid model. Its basic feature is that the control intelligence of the optical transport layer is transferred to the IP layer, which implements end-to-end control. In this case, the optical transmission network and the IP network form an integrated network, and the unified control plane maintains a single topology, optical Switches and IP routers can freely exchange all information and run the same routing and signaling protocols to achieve integrated management and traffic engineering. But its disadvantage is also obvious, that is, a large amount of state and control information must be exchanged between the optical layer and the IP layer.
3.1 GMPLS label and label exchange path
In order to control optical networks, GMPLS not only supports traditional group switching, but also supports time division switching, wavelength switching, and optical fiber switching. This determines that GMPLS is very different from MPLS, it is mainly manifested in the following aspects:
· MPLS labels have a very large space, while wavelength and time-division channels are very limited.
· Mpls lsp can be allocated with continuous bandwidth, while optical channels and time-division channels can only be allocated with limited discrete bandwidth.
· If multiple parallel optical fibers exist between two nodes, GMPLS must also support Optical Fiber switching.
3.1.1 GMPLS label
In order to support circuit switching, SDH) and optical switching, including LSC and FSC, GMPLS has designed a special label format. Labels should support the identification of optical fiber, wave band, wavelength and even time slot. Taking the TLV format of CR-LDP as an example, its label item should contain four fields: LPT, LSP-ENC, G-PID and link identifier. Here, the LPT field refers to the link protection type, the LSP-ENC field refers to the LSP encoding type, defines the OC-nSONET), STS-nSDH), GigE, 10 GigE, DS1 ~ DS4, E1 ~ E4, J3, J4, VT, optical wavelength, band, and other types. The G-PID field is the generic Net Load identifier, indicating the Net Load type carried by LSP, using the standard Ethernet Net Load type, set by the inbound node, for use by the outbound node, the intermediate node only performs transparent transfer. The link Id field identifies the link that receives the tag request. It takes effect only locally between adjacent nodes. The tag length and format vary depending on the application environment. For example, in wavelength label switching, the port/wavelength label is 32bit, indicating the optical fiber or port or wavelength used, unlike the traditional tag, there are no experimental bit, tag stack bottom, TTL, and other domains. However, like the traditional tag, it is only effective locally between adjacent nodes. The tag value can be determined dynamically by manual configuration or protocol.
3.1.2 hierarchical label exchange path of GMPLS
To support optical networks, GMPLS needs to introduce a new concept-hierarchical tag switching path. The hierarchical meaning is for the reuse capability of LSP. The higher the reuse capability, the higher the LSP level. 1 shows that LSP1, LSP2, LSP3, and LSP4 have a low-to-high nested relationship. LSP1 is at the lowest layer. Its start and end devices are vrouters for network interfaces with group switching capabilities ); LSP1 and other LSP with grouped transmission capabilities can be aggregated into LSP2. LSP2 is on the second layer, and its start and end devices are network interfaces with Time Slot Switching capabilities, the main types are SDH/SONET, TDM, or ADM interfaces. Similarly, LSP2 can be aggregated into LSP3 with other LSPs with Time Slot transmission capabilities, the starting end and end devices of LSP3 are optical crossover devices, OXC) in a network with wavelength switching capability; LSP4 is at the highest level in a network with Optical Fiber switching capability.
LSP label switching path) after layering, the benefits are obvious. First, the wavelength and time slot channels can be used very economically through route aggregation between different layers to solve the problem of limited wavelength and time slot channels. Second, it solves the problem that only a limited number of discrete bandwidth values can be allocated to optical channels and time-division channels. For example, before using a layer-by-layer LSP, an independent, very large discrete bandwidth, such as 2 Gbit/s, is required for LSPs that pass through the optical network 100 Mbit/s ). After the layered structure is adopted, each wavelength channel becomes an aggregation route. A large number of LSPs can share a 2 Gbit/s optical channel.
3.1.3 establish hierarchical LSP
This part will explain the hierarchical LSP establishment process, assuming LSP1 is a line Supporting 500 Mbit/s packet transmission, LSP2 is a STS-12c SDH line, LSP3 is a OC-192 line, LSP4 is a line that supports WDM.
The following discussion is based on the extended RSVP-TE signaling defined in GMPLS. The original RSVP uses two types of signaling. One is the PATH message, which is the request information sent from the initiator to the receiving end. It mainly contains parameters for the business flow Description and classification. The other is RESV information, which contains the resource parameters that describe the acceptor reserves. To support MPLS, you need to add a label object to the RESV information. The simple working principle is: When an LSR needs to send RESV information to an RSVP stream, it generates a new tag and writes it to the label column of the forwarding table and the RSVP information to be sent. When the upstream adjacent LSR receives this information, it writes the tags in the RESV information to the output tag column of the forwarding table and generates a new tag, and write it into the label column of the forwarding table and the RSVP information to be sent, and then the information is transmitted to the upstream adjacent LSR. When the RESV information reaches the origin, an LSP that guarantees QoS is established.
The hierarchical LSP creation process is as follows:
1) A path request packet Path1 about creating LSP1 is generated in R0. This packet is forwarded to an edge node of R1 group exchange network ).
2) When R1 receives this packet, it will trigger the path request packet Path2 to establish LSP2R1 to R7). This packet is forwarded to S2, this process continues until the path request packet Path4 of LSP4 is generated.
3) When Path4 arrives at O5, O5 returns the Resv information along the original route. When the Resv information reaches O3, LSP4 is successfully established. At this time, Path3 packets can be transmitted from LSP4 to O5, o5 then forwards the message to S6, and S6 sends the tag ing message along the original radial S2, and LSP3 is then created. This process repeats until LSP1 is successfully created.