From the perspective of technological development, using ATM to build a broadband Internet backbone network usually goes through three major technical stages. This article analyzes the main technologies involved in these three phases and takes the next-generation Internet trial network VBNS in the United States as an example.
1 Introduction
As we all know, Internet has experienced explosive growth in the world in recent years. According to statistics, the number of Internet users has reached 64 million, with an average increase of 0.95 million users per month. On average, there is an Internet access every 30 minutes. The sharp increase in the number of users and the popularity of WWW applications with high bandwidth requirements have led to a sustained increase in Internet traffic, resulting in network congestion. To expand Internet bandwidth, more and more Internet Service Providers begin to use ATM to construct a broadband Internet backbone network, and lay the foundation for providing comprehensive services on the Internet in the future, how can I use ATM to build a broadband Internet backbone network? This is a matter of public concern. From the perspective of technological development, the establishment of TCP/IP backbone networks using ATM generally goes through three major technical stages. This article will analyze the main technologies involved in these three phases, and take the next generation Internet trial Network vBNSvery high Bandwidth Network Service in the United States as an example.
2 Stage 1: edge routing)
This is the current stage of technology. There are three main protocols or specifications used: PVC-based transparent bridging PVC bridging), and SVC-Based Classical Over ATMRFC1577) and LAN simulation LANE ). The first two protocols are recommended by the Internet technical organization IETFInternet Engineering Task Force, and the last one is developed by the atm forum.
2.1 PVC-based transparent bridging PVC bridging)
Transparent bridging (transparent bridging) is a technology that connects local or remote LAN network segments to form a logical subnet. In an environment with multiple LAN segments, the transparent bridge) network device can filter LAN frames, only transmit frames to the destination CIDR block rather than to other CIDR blocks. It checks the source MACMedia Access Control address of each LAN frame to determine the network segment of the device that sends the frame and remembers it. Later, bridge) transmits the frame to the corresponding network segment based on the target MAC address of each LAN frame.
Figure 1 shows that two network devices with ATM ports use PVC-based transparent bridging PVC bridging to interconnect two remote LAN network segments. It should be emphasized that each LAN frame must be specially encapsulated with encapsulation before being transmitted on the ATM link. The Protocol followed by this encapsulation is RFC1483, which defines how packets of Ethernet, licensing ring, and FDDI are encapsulated into ATM cells, so as to realize transmission of traditional LAN frames over ATM.
It is very easy to establish a TCP/IP Backbone Network on an ATM using the PVC bridging technology. It establishes an atm pvc connection between remote network devices and implements Layer 2 interconnection through transparent bridging. This networking method is simple and efficient. It is interconnected on the second layer, that is, the data link layer, to achieve high throughput. The disadvantage is that the number of PVC increases exponentially as the number of backbone nodes increases, the networking is not flexible enough.
2.2 CLassical IP Over atm clip)
CLIPRFC1577) is a IP-OVER-ATM Protocol recommended by IETF Based on the ATM Forum UNI3.0/3.1 signaling standard. Its principle is similar to the IP-MAC Address Resolution Protocol ARP: Address Resolution Protocol in TCP/IP network, IP-ATM Address Resolution Server ARP Server needs to be established in the Network), ARP Server can be implemented in ATM device or through Server software. CLIP client that belongs to the same IP subnet) sends a IP-ATM address resolution request to its ARP Server, ARP Server Returns the ATM address of the destination terminal to be accessed by the CLIP terminal, the source terminal establishes a SVC connection with the destination terminal through the ATM UNI3.0/3.1 signaling to achieve communication with the destination terminal. Similar to PVC bridging, each IP packet must be specially encapsulated before being transmitted on the ATM link. The Protocol followed is RFC1483.
Compared with PVC bridging, CLIP is implemented at the IP network layer, which shields the heterogeneity of the underlying physical network and supports SVC, which can also be PVC. Therefore, the networking is more flexible, at the same time, high efficiency is maintained.
2.3 LAN simulation LANE)
The specification is developed by the ATM Forum, which defines how to simulate an ATM network into a second layer logic with a large network or a ring-based network segment-simulation LAN emulated LAN ). In the simulated LAN environment, see figure 2), each LANE's terminal LEC-LANE Client) it can be a host with an ATM Nic, can also be a LAN switch with an ATM port Layer 2 LAN switch) when joining the simulation LAN, first access LANE Configuration Server LECS-LANE Configuration Server ), LECS records the ATM address of the simulated LAN to which the LANE terminal belongs and Its LANE server LES. After obtaining the LES address, LEC establishes a communication with it and registers it as part of the simulation LAN on LES. To simulate the traditional lan cidr blocks, a public IP domain covering all LANE terminals is also established. The LANE Specification defines a Broadcast/Unknown terminal Server BUS: broadcast and Unknown Server), which transmits the Broadcast package broadcast packet and multicast packet to each member of the simulation LAN, it is also responsible for processing packets with unknown ATM addresses of the destination terminal. Similar to the ARP Server in the CLIP environment, LECS, LES, and BUS can be implemented either on an ATM device or through server software.
In a simulated LAN, an IP terminal wants to communicate with another IP terminal. It first sends a resolution request to the BUS for the IP address of the destination terminal, BUS forwards the request to all terminals in the simulated lan. After receiving the request, the destination terminal provides its own MAC address. After obtaining the MAC address of the destination terminal, the source terminal sends a resolution request to LES for the MAC address of the destination terminal, and LES responds to the ATM address of the destination terminal. As a result, the source terminal obtains the destination terminal's ATM address, which can establish a SVC connection with the destination terminal through the ATM signaling.
Using LAN simulation technology to build a TCP/IP Backbone Network on an ATM, LECS, LES, and BUS servers must be set up on the ATM network to form a virtual network together with relevant network devices, the network device establishes a connection through SVC to interconnect the same LAN segment. Compared with CLIP, LANE is implemented at the second layer, that is, the data link layer level. The implementation mechanism is complex and requires a large amount of network resources and is less efficient, however, it can achieve intercommunication between FDDI, Ethernet, and card ring. It is also compatible with the MPOAMultiProtocol over ATM specifications to be released at the ATM Forum, allowing users to smoothly transition to the MPOA environment.
2.4 application instance-vBNS Network
In 1994, the National Foundation for Science and Technology (NFS) invested $50 million to entrust companies like the United States to research the next-generation high-speed Internet backbone network. In 1995, the company established vBNS, the next-generation Internet trial network spanning the United States. See Figure 3. vBNS uses ATM as the backbone of the TCP/IP network, and the switches are connected by SONET OC-3155Mbit/s. The Network is connected to five super computing Centers (SCC) and four large Internet Access points (NAP-Network Access Points) through CISCO and NetStar routers, and through NAP or leased line to DS-345 Mbit/s) or OC-3155Mbii/s) with other education and research units. All these connections support CP/IP. A point-to-point atm pvc is established between routers connected to an ATM network to form a full-mesh connection. Transparent bridging is used to achieve layer-2 interconnection. On this basis, OSPF routing protocol is run. Because PVC bridging is used to establish a TCP/IP Backbone Network on an ATM, vBNS breaks through the bandwidth bottleneck of the traditional Internet backbone network and achieves high efficiency. Performance Tests on hosts in two supercomputing centers show that the peak rate of UDP packet User datasync Protocol packet reaches 133 Mbit/s, it is very close to the theoretical peak rate of the IP packet transmitted on the ATM OC-3 135 Mbit/s ). In order to reduce network complexity and improve reliability, MCI also plans to change the PVC mesh connection between routers to SVC in 1997.
3 Stage 2: direct routing)
3.1 edge routing Defects
Although EDGE routing technology can be used to transmit TCP/IP over an ATM network, this solution is not satisfactory. First, both PVC bridging, CLIP, and LANE construct the second-Layer Virtual Network Layer 2 Virtual LAN on the ATM. The terminal in the Virtual network or the LAN segment in the same traditional LAN segment) or belong to the same IP subnet. Communication between different virtual networks still needs to pass through the traditional router. Generally, the router is located at the edge of the ATM network. Therefore, these technologies are collectively referred to as edge routing technologies ). Even for two devices directly connected to the ATM, if they belong to different IP subnets, the source terminal must first establish a connection with the vro and then communicate with the destination terminal through a route, therefore, the router becomes the bottleneck of TCP/IP network expansion and performance, and IP transmission introduces latency. Second, the establishment of a TCP/IP Backbone Network on an ATM using the second-layer virtual network will introduce broadcast on the Internet, consuming the CPU resources of the device and wasting the bandwidth of the ATM Backbone Network. Third, the edge routing solution does not take advantage of the advantages of ATM: multimedia communication support and Quality of Service (QoS) assurance, even if the High-level applications carried by TCP/IP need to put forward service quality requirements for the transmission platform, edge routing technology does not define how to deliver the High-level service quality requirements to the ATM layer. Therefore, it is necessary to seek further solutions.
3.2 virtual Router Viriual Router) -- MPOA Model
The second stage mainly solves the problem of direct routing of TCP/IP over the ATM network. The main idea is to adopt the customer/Server mode. The edge device asks the Routing Server for a route, and the Routing Server gives an answer to the egress, the edge device establishes a SVC connection with the destination terminal based on the answer. Typical examples of this model include Ipsilon's IP Switching, CISCO's TAG Switching, and MP0A specifications developed by the MPOA Working Group of the ATM Forum. MPOA is introduced here.
An MPOA Model 4) mainly includes the following devices:
Edge device (edge devices)-it is a smart switch. One end is connected to an ATM, and the other end is connected to a traditional LAN. It performs Packet Exchange between a traditional LAN segment and an ATM based on the destination terminal's network address, such as an IP address.
ATM host-attached hosts) -- it is a host with an ATM Nic, which supports the MPOA protocol, this allows the host to communicate with a traditional LAN connected to other ATM hosts or edge devices. Edge devices and ATM hosts are collectively referred to as end-points ).
Routing server (Route server)-It is responsible for maintaining the network addresses of all terminals of its services, such as IP addresses), MAC addresses, and ATM address information. The terminal uses this information to establish a direct connection with other terminals. It also runs RIP, OSPF, or a future I-PNNI routing protocol Integrated-PNNI, see below) exchange routing information with a traditional router. A Routing Server can be a separate product or be implemented in an existing vro or vswitch. -- The core of the MPOA model is the concept of virtual routers. An MPOA model running on an ATM network is equivalent to a multi-layer router: an edge device receives data transmitted by a traditional LAN and can be viewed as a router interface card. An ATM network connects edge devices, it can be viewed as the backplane of A vro. The Routing Server is responsible for address and route resolution, information flow analysis and control, and can be viewed as the control processor of the vro ).
MPOA is used to establish a TCP/IP Backbone Network on an ATM. It is required to set up routing servers, terminal edge devices, and ATM hosts on the Internet. MPOA or LANE protocols should be supported, in this case, the terminal must communicate with other IP subnets through a traditional router or a terminal that supports MPOA. When the terminal receives an IP packet, it first checks the destination address of the IP packet to determine how to forward the packet. If the destination address is still in the same IP subnet, it uses the LANE Protocol to parse the ATM address of the destination terminal and then establishes an ATM connection with the destination terminal. If it is not in the same subnet, it initiates an IP address resolution request to the Routing Server, the routing server responds to the ATM address of the destination terminal, so the source terminal can establish a SVC connection with the destination terminal. Obviously, through MPOA, direct routing can be achieved between ATM network devices, which breaks through the expansion and performance constraints caused by edge routes and reduces unnecessary broadcast traffic on the Internet, this laid the foundation for the next step to use the service quality characteristics of ATM to support integrated services on the Internet.
4 Stage 3: shared routing)
In the third stage, the TCP/IP network adopts the same routing protocol as the ATM network, which makes the network organization more effective and can provide comprehensive Internet service by using the ATM features. What technologies and specifications are used in this phase are still being studied. Here we will mainly introduce the PNNI extended Routing PAR-PNNI Augmented Routing discussed in the ATM Forum) and the Integrated PNNI Routing I-PNNI-Integrated PNNI Routing) solution.
The existing scheme that supports TCP/IP Transmission over ATM, such as LANE and MPOA) is based on layered routing (layeredrouting, the ATM network provides a set of point-to-point circuit connections. On this basis, the router implements OSPF, RIP, and other routing protocols. IP routing is not used in the ATM Routing. This hierarchical view has limitations. If the network administrator does not determine the connection relationship in advance, the two routers connected through the ATM network are directly connected to the ATM or connected to the ATM through the edge device), the other party cannot be found and the route connection is automatically established. In addition, the PNNI protocol developed by the ATM Forum defines the signaling and routing protocols between switches in the ATM network. Through this protocol, the ATM switch can meet a series of selection policies, including service quality requirements) to determine the route and automatically find the best route. PNNI also hierarchically divides the ATM network by routing domain, so that the network can be expanded to a large scale of thousands of switches ). If layer-based routing is used to establish a TCP/IP Backbone Network on an ATM, these advantages of PNNI cannot be used.
In order to make up for the above shortcomings and establish a TCP/IP backbone network more effectively on the ATM, people have proposed a PNNI Extension Route and a comprehensive PNNI Route Scheme. The first step is to establish a PNNI extended route to allow the routers connected to the ATM to participate in the route of the ATM network, the PNNI protocol is also executed on the ATM port to exchange route information with the ATM network, so that the router interface can more effectively use the atm svc connection to support high-level applications, including support for service quality. More importantly, the PAR allows the router to transmit the relevant IP information through the PNNI Protocol, so that the router can automatically discover other routers connected to the ATM network and establish a SVC connection with them without the need of the network administrator's preset settings, then, implement OSPF and other routing protocols to establish a TCP/IP backbone network. The above mechanism is transparent to the ATM switch, and the switch does not need to be upgraded. It only needs to treat the PAR router as a general ATM switch.
The second step is to establish a comprehensive PNNI route. In the PAR environment, the PAR router must be directly connected to the ATM, and the PNNI and traditional routing protocols such as OSPF and RIP must be run at the same time; in the I-PNNI environment, all routers, including routers not directly connected to the ATM) run only one new routing protocol-I-PNNI, I-PNNI and PNNI can interwork, can also run on the ATM switch. By running the I-PNNI on the router, all the routers are involved in the routing of the ATM network, the traditional LAN and the ATM network to achieve seamless connection. A broadband Internet network can use the advantages of the ATM and PNNI routing protocols to provide comprehensive service services with guaranteed service quality and establish a layered network structure, the TCP/IP network can be expanded to a large scale.
It should be noted that the I-PNNI and MPOA are compatible with MP0A but only define the concept of virtual router and the functions of each part, the Routing Server is responsible for routing calculation, and the terminal is responsible for packet transmission. MPOA does not define the routing protocol used by the Routing Server to exchange routing information with the traditional router, The I-PNNI routing protocol can be used as a choice.
5. Summary
To sum up, using an ATM to build a TCP/IP network usually goes through three technical stages: edge routing, direct routing, and shared routing. At present, the edge routing technology is relatively mature, and some technical solutions have been introduced for direct routing. It may take some time for anyone to become the mainstream. The sharing routing technology is still in the research stage. When using ATM to build a broadband Internet, network operators should not only consider the efficiency and flexibility of the current network, but also take into account the smooth transition of China Telecom to the next stage)
