Basic Principles of MPLS technology (1)

1. MPLS Technology

The proposal of Multi-Protocol Label Switching (MPLS) technology the rapid growth of Internet business volume and the emergence of broadband technology poses a severe challenge to Internet service provider (ISP) network bandwidth. This challenge is not only the requirement for high bandwidth, but also the requirement for the traditional route switching mode based on the current Internet.
To establish such a new generation Routing System with good service quality and scalability support, Each router needs to maintain large-capacity routing information and establish a hierarchical structure of routing information; in addition to enhancing the router's forwarding performance to IP packet groups, you also need to add route support for multi-object broadcast to provide a hierarchical routing information structure. In addition, in the future, the routing system must have flexible adaptability to meet various new requirements that may arise in the future.
From the perspective of Internet maintenance and application, how to implement reasonable and simple traffic control (TrafficEngineering) for the Internet and Implement Virtual Private Network (VPN) based on IP Services ), ensuring IP-level service quality (IP-levelQoS) poses challenges to the ISP Backbone Network Based on the traditional network topology and IP routing system.
2. Technical Principles of MPLS
2.1MPLS protocol and functions
(1) concept of routing and switching
Before describing the MPLS technology in detail, we should first review several concepts related to the exchange technology.
The routing protocol (such as RIP and OSPF) is a mechanism that makes every device in the network know that when a group is sent to its destination, the Next Hop level for transmitting this group. A router uses the routing protocol to build a route table. When a router receives a group and must make a forwarding decision, the router uses the destination IP address in the group as the Index to search for the route table, obtain the address of the next-hop machine using a specific algorithm. The construction of Route tables and the search for them during forwarding are basically two independent operations.
The exchange concept is usually used to describe the data transfer from an input port in a device to an output port. This transfer is generally based on Layer 2 (such as ATMVPI/VCI) information.
The control component creates and maintains a route forwarding table for a node ). It works with the control components of other nodes to continuously and correctly exchange distributed routing information and create a forwarding table locally. Standard routing protocols (such as OSPF, BGP, and RIP) are used to exchange routing information between control components.
The forwarding part performs the group forwarding function. It uses forwarding table, address shared by the group, and a series of local operations for forwarding and decision. In traditional routers, the longest matching algorithm compares the destination address in the group with the entries in the forwarding table until an optimal match is obtained. More importantly, repeat this operation from the source node to the destination node. In a logo exchange router, the (best match) Logo exchange algorithm uses the logo of the group and the logo-based forwarding table to obtain a new logo and output port for the group.
The route forwarding table contains several items, which provide information to the forwarding part and perform its switching function. The forwarding table must associate each group with one entry (the traditional entry is the destination address) to provide guidance for the next route entry of the group.
The same type of forwarding (FEC) defines such a group. From the perspective of forwarding behavior, they all have the same forwarding attribute. A fec is a group of single-object broadcast groups, and its destination addresses all match an IP address prefix. Another type of FEC is a group with the same source and destination addresses. FEC can be defined at different levels.
The label is relatively short and has a fixed length and no structure identifier. It can be used in the forwarding process. A flag is associated with a FEC through a binding operation. Under normal circumstances, a single data link only has local significance and does not have global significance. In an ATM environment, they are equivalent to their VPI/VCI. Because ATM uses a fixed short area for exchange, we can believe that the flag exchange can be an effective solution for IPoverATM applications. Under an event-driven system, the flag is bound to FEC, which has some significance. Such events can be divided into the following two types:
One is data-driven binding, that is, binding when data streams are generated. The flag binding is created only when required. There are only a few items in the forwarding table. Indicates that data streams are allocated to different IP addresses. In an ATM network environment, it requires a large amount of virtual circuit resources and is not easy to expand.
The other is topology-driven binding. When the control plane is activated, it is irrelevant to the generation of data streams. Flag binding may be related to route updates or receipt of RSVP messages. Topology-driven binding is easier to expand than data-driven binding, so it is used in MPLS.
(2) Logo exchange and forwarding parts
There are several ways to bind a flag to a group. Some networks can embed the flag into the header of the link layer (ATMVCI/VPI, And the DLCI of Frame Relay ). Sometimes it can be embedded into the data link header and the Data Link Protocol Data Unit (PDU) A small sign header (such as located between the second-level header end and the third-level data load) is called "Shim ".
This flag information can be carried at the link layer. The "Shim" structure can be used on Ethernet, 802.3, or point-to-point (PPP) links. One of them is a single-camera broadcast, the other is multi-object broadcast (Multicast ). Each flag is 4 bytes.
On the edge of an MPLS backbone network, the boundary LSR classifies incoming unlabeled groups (normally) by their IP address headers (Classification) and determines the forwarding, in this way, the IP Group is marked with a corresponding identifier on the boundary LSR and transmitted to the next hop at the destination address.
In the subsequent exchange process, the fixed-length logo generated by LSR replaces the IP grouping header, greatly simplifying the subsequent node processing operations. Later nodes use this flag for forwarding decisions. In general, the value of the flag changes after being exchanged in each LSR. This is the flag forwarding.
If a group comes out of the MPLS backbone network and the egress boundary LSR finds that their forwarding direction is a non-flag interface, it simply removes the flag from the group. The most important advantage of this kind of logo-based forwarding is that only one forwarding algorithm is required for multiple exchange types. You can use hardware to achieve a very high forwarding speed.
(3) Mark switching control components
A flag is appended to the group by an Upstream LSR (Upstream LSR) node of the Flag exchange path (LSP). The Downstream LSR (Downstream LSR) receives the flag and then makes a decision, this is done by the control part of the sign exchange. It uses the items in the logo forwarding table as a guide.
In addition to basic table creation and maintenance, the mark exchange control component is also responsible for routing distribution between LSR in a continuous manner and generating the information into a forwarding table. The flag switching control components include all traditional routing protocols (such as OSPF, BGP, and PIM ). These routing protocols provide LSR with the ing between FEC and the next hop address.
Distribution)
The entries in the logo exchange forwarding table should at least provide the output port information and the next new logo. Of course, more information can be included. For example, it can generate an output queue principle for the switched group. The input group must have a unique entry corresponding to the entry in the forwarding table.
Each assigned flag must be associated with an entry in the forwarding table. This binding can be executed on the local LSR or on the remote LSR. Currently, the MPLS version uses downstream binding. In this case, the local association flag is used as the entry group flag, and the remote Association flag is used as the output flag. Another method is upstream binding, which is opposite to downstream binding. In MPLS technology, the forwarding table is also called the mark forwarding information library (LFIB). Each entry of LFIB includes the input mark, output mark, input interface, and output port MAC address, search for items with the input flag. In addition, LFIB can exist on either a logo exchange router or an interface.
(4) Mark exchange router (LSR)
MPLS devices can be divided into border signs exchange routers and intermediate signs exchange routers according to their locations in the MPLS routing network. In addition to adding or removing the Group logo, the boundary LSR is also responsible for classifying traffic. In addition to the destination address, there are many other factors in the allocation of a flag. The boundary LSR determines whether the traffic is a persistent stream. It adopts management policies and access control policies, and aggregates common business flows into large data streams when possible. These are all functions required at the IP and MPLS boundary. Therefore, the capability of the boundary LSR will be the key to the success of the entire marking exchange environment. This is also a management and control point for service providers.
(5) Relationship between MPLS and ATM protocols
MPLS is a common forwarding algorithm. In combination with ATM technology, MPLS uses the user plane of ATM ), using the VPI/VCI of an ATM as its flag; the control function of mpls is based on the dynamic routing protocols (such as IS-IS, OSPF, BGP, and PIM) at the network layer) and the flag Allocation Protocol (LDP) to replace the traditional control plane of the ATM, to complete the control function of the entire MPLS network.
3. MPLS features

Traditional IP data forwarding is based on the hop-by-hop mode. Each router that forwards data needs to find the route table based on the destination IP address of the IP address header to obtain the next hop egress, this is a tedious and inefficient task, mainly because of two reasons: 1. Some route queries must perform multiple searches on the route table, which is called recursive search; 2. Because route matching follows the longest Match Principle, almost all vro switching engines must be implemented using software, the exchange engine implemented by software cannot compete with the switch engine implemented by hardware on the ATM switch in terms of efficiency.
MPLS technology is proposed to better combine IP and ATM high-speed exchange technology, give full play to the advantages of the two, make full use of the various resources of the current ATM network, implement fast forwarding and exchange of IP groups, expand the traditional dynamic IP Routing, and implement control-Based Dynamic Routing (Constraint-Based Routing) implements IP service traffic control, BGP/mpls vpn for virtual private networks, and IP-level service quality (IP Cos ).
Compared with other technologies, the Multi-Protocol Label Switching MPLS technology has three features:
(1) MPLS switching is different from traditional IP routing. It is based on an explicit route switching (explicit routing) and source address routing mode. 2) The label used in MPLS does not have a fixed format and changes with the changes of the lower-layer media. For ATM media, the label is the VCI/VPI of the ATM, and the frame relay is the DLCI, for X.25, It is LCN. 3) MPLS route control management is a network topology-oriented implementation that drives topo-driven. Only when the entire network topology changes, the MPLS route forwarding table changes without changing an application service or a workstation in the network.
In the MPLS switching technology based on ATM, the prototype of "label packaging" is very similar to that of the forwarding cells of the ATM switch. from another perspective, in a label switching environment, the ATM switch will be more like a fast router. A typical method of MPLS is to assign a label with a specific meaning to the routing prefix in the 3rd-layer route table. This topology-driven) Flag allocation technology is different from other flow-driven (flow-driven) allocation technologies. The labels assigned in MPLS only change with the change of the routing prefix, the frequency of change is much lower. Obviously, this technology is highly scalable because it is independent of data streams and uses destination addresses. To some extent, the flag exchange function is similar to the data link connection identifier (DLCI) used by Frame Relay for high-speed switching. When the marked Group finally reaches the exit of the marked network, it is removed and forwarded by the traditional 3rd-layer route.