ATM Tutorial: Support for IP addresses in an ATM Network

In the past decade, ATM has become an important technology for next-generation networks. It provides unprecedented scalability and cost-effectiveness, as well as support for real-time and multimedia services in the future. In the future information system, ATM will play an important role. However, the current information system, LAN and WAN, is built on network layer protocols such as IP, IPX, and AppleTalk. Therefore, the key to the success of ATM and the development of Internet is the existing network technology and the interoperation of ATM. The key to achieving this goal is the same network layer protocol, such as IP, IPX, it is also applied to existing networks and ATMs because it is a network layer task to provide a unified network perspective for high-level protocols and applications. So far, there have been a variety of methods to run IP addresses on the ATM, such: in the ATM Forum, LANE, MPOA, IETF's CLIP, NHRP, Ipsilon network, and Cisco's tag exchange are described one by one.
I. Introduction
ATM and the existing protocol systems, especially the IP and IPX protocols at the network layer, coexist for a long time, how to implement the existing network protocol and ATM on a single network, and how to connect the ATM with the traditional network is a topic of researchers, designers and practitioners. However, ATM and IP address come from different technical groups and foundations and have their own applications. The purpose of an IP address is to send packets to the destination in an uncertain state. It is non-connected and has no guarantee of service quality. The purpose of an ATM is to provide integrated services with guaranteed performance and is connection-oriented, fast fixed-length cell switching. The huge difference between ATM and IP makes it difficult to effectively integrate the two.
There are two different models for supporting IP addresses in an ATM network. These two models view the relationship between the ATM protocol layer and IP addresses from different perspectives.
The first is the peer-to-peer model. In essence, the ATM layer is regarded as the Peer-to-Peer layer of IP addresses. This model recommends that the same IP address scheme be used in the ATM network as in the IP-based network, therefore, the ATM endpoint is identified by the IP address. The ATM signaling carries such an address, and the routing of the ATM signaling also enables the existing network layer routing protocol. Because existing routing protocols are used, the Peer-to-Peer Model eliminates the need to develop new ATM routes. The peer-to-peer model simplifies terminal system address management while greatly increasing the complexity of the ATM switch, because the ATM switch must have the multi-protocol router function, supports existing address schemes and routing protocols. In addition, the existing routing protocols are developed based on the current LAN and WAN, and cannot be well mapped to the ATM and the service quality characteristics of the ATM.
In the current solution, IP address switching and tag switching are based on the peer-to-peer model.
Another model is called Subnet or coverage model, which separates the ATM layer from the existing protocol and defines a new address system, that is, the existing protocol will run on the ATM. This overwrite model requires the definition of a new address system and related routing protocols. All ATM systems need to be assigned both the ATM address and the high-level protocol address to be supported. The ATM address space is logically separated from the address space of the top-level protocol, without any correlation. Therefore, all protocols running on the ATM subnet need some ATM Address Resolution Protocol to map high-level protocols such as IP addresses to the corresponding ATM address. This method of separating ATM from high-level protocols allows independent development, which is very important for practical engineering.
In the current solution, LANE, MPOA, and CLIP are based on the coverage model.
Ii. LANE
1. How to run an IP address on a traditional LAN?
In a traditional LAN, when the source host wants to send a group to the target host in the same subnet, it checks its ARP cache to see if it knows the hardware address (MAC address) associated with the IP address of the target host. if it knows, send the group with the IP address and MAC address of the target host.
If the target MAC address is unknown, the source host sends an ARP request group. The ARP request is a local broadcast group that receives all hosts in the subnet. After the target host recognizes its IP address, in the ARP response group, the source host receives the ARP response and stores it in its own ARP table, now, the source host can send a group containing the correct destination IP address and MAC address.
2. What functions must ATMLAN simulate?
(1) because the traditional LAN is a media shared network, it is easy to provide broadcast services and implement ARP. the ATM network must imitate this function, which is implemented by BUS broadcast and unknown servers.
(2) In general, each host in a traditional LAN has its MAC address and IP address. In addition to its ATM address, the host directly connected to the ATM network must also have a MAC address and IP address.
(3) The ATM host must provide the same services as the interface services provided by the MAC protocol to the network layer protocol, such as NDIS or ODI-Type Driver interfaces.
3. How does LANE work?
As the name suggests, LANE is used to simulate a LAN on an ATM network. The LANE Protocol defines a mechanism to simulate a 802.5 Ethernet or ring network. The LANE Protocol defines the same interface as the service provided by the existing LAN to the network layer. The data transmitted in the ATM network is encapsulated in the corresponding LANMAC group format.
Each ELAN (EmulatedLAN) consists of a group of LANE customers (LEC) and LANE Services. LEC can also be a bridge and router used as an ATM host proxy. The LE Service consists of three different functional entities: the LAN simulation configuration server (LECS), the LAN server (LES), and the BUS. These three service entities can exist separately, but it is usually located on the same device. For example, LES can be located on an ATM switch, router, bridge, and workstation.
Below are the steps for the workstation in LANE to communicate with another workstation:
(1) initialization
LEC needs to know the lecs atm address and establish a connection with it. This is done through ILMI or the well-known LECS address, during this process, LEC can establish a bidirectional configuration with the manually configured LECS address to go directly to VCC. In this process, LEC will obtain the ATM address of the elan les.
(2) Registration
This is the mechanism that LEC provides the address information for LES, such as the MAC address. In this process, a pair of connections will be established between LEC and LES, that is, two-way point-to-point control direct to VCC, and one-way point-to-multi-point control distribution VCC.
(3) Address Resolution
This is how LEC learns the ATM address of the target site from LES. It is implemented by the ATM Address Resolution Protocol and allows LEC to establish data to connect to VCC to transmit frames. At this time, two-way point-to-point Multicast Transmission VCC and one-way point-to-Multi-Point multicast forwarding VCC are established between LEC and BUS.
(4) Data Transmission
When the source site and target site are waiting for the establishment of data to reach the transitional period of VCC, the BUS can forward the frame to all LEC in the ELAN. When the data goes directly to the establishment of VCC, communication switches from the original route (BUS) to the new route. To ensure the frame sequence, the Information Clearing protocol (flushmessageprotocol) is used to notify the BUS: when a new route is used to transmit frames, the request is cleared and sent to the BUS and forwarded to all LEC in the ELAN, and no frames are transmitted through the old BUS route ), all frames are sent to the target site through data direct to the VCC new route.
It should be noted that in the above process described in the ATM Forum specification, resolution from IP address to MAC address is not mentioned. The entire process of communication between a traditional LAN host and an ATM host is as follows:
(1) To determine the MAC address of the target site, the source host broadcasts an ARP request containing an IP address, which is a standard process of any IP network. The ARP request reaches the LAN/ATM Bridge on the traditional LAN.
(2) LEC on the LAN/ATM bridge forwards the broadcast group VCC to the BUS through multicast, and the BUS forwards the VCC to all Members in the ELAN through multicast to send ARP requests.
(3) the target site receives an ARP request and identifies its own IP address. In response, it puts its MAC address in the ARP response. Because it is not directly connected to the LAN/ATM bridge to VCC, the LEC of the target site sends the ARP response to the BUS through multicast, the BUS sends a VCC to the LAN/ATM bridge through multicast.
(4) LAN/ATM bridges send ARP responses to the source host through the traditional LAN.
(5) The source host has the MAC address of the target site and starts to transmit data over the LAN.
(6) The Bridge sends VCC packets to the BUS through multicast, and the BUS forwards the packets to the target site.
(7) At the same time, LEC on the LAN/ATM bridge sends a LE-ARP request to LES through direct control to the VCC, asking the ATM address corresponding to the MAC address of the destination site, if LES does not have this ing, A LE-ARP request is sent to all LEC through control distribution VCC, after receiving the request, the destination site LEC puts its ATM address in the LE-ARP to respond and sends it back to LES through control to VCC.
(8) The source LEC receives a LE-ARP response from LES through direct control to VCC, extracts the ATM address and establishes data between the source and destination to VCC.
(9) after data is directly transmitted to the VCC, packets sent from the bridge are transmitted directly to the VCC to replace the BUS.
4. advantages and limitations of LANE
Because LANE provides the same service interface as the driver provided by the existing MAC protocol to the network layer, no need to change the driver, which will accelerate the development and application of ATM. However, the function of LANE is to make the ATM feature transparent to the high-level protocol, so it makes the high-level protocol unable to take advantage of the inherent advantages of ATM, especially its service quality assurance. The new LANE2.0 version provides local management service quality for inter-system communication at the ATM end. The Protocol provides a mechanism to determine whether to support the expected service quality. Each locally defined service quality can contain information to indicate whether VCC established with this service quality can be shared by other protocols or applications.
Although LANE provides an effective way to bridge between subnets in an ATM network, services between subnets still need to be forwarded through routers. Therefore, an ATM router may become a bottleneck, the MPOA mentioned below will solve the problem of inter-subnet communication efficiency.
　
Iii. CLIP (ClassicalIPoverATM)
1. Principle
To run IP addresses on an ATM network, IETF adopts the concept of a logical independent IP subnet (LIS. Like a common IP subnet, a LIS contains a group of IP nodes connected to a single ATM network, such as a host or router. They belong to the same IP subnet. ATMLIS acts like a traditional IP subnet. To resolve the Node Address in LIS, each LIS provides an ATMARP server. All nodes in the LIS (LIS customers) configured with the ATM address of the ATMARP server. When a node in LIS appears, it first establishes a connection with the ATMARP server. Once the ATMARP server detects a connection to a new LIS customer, it sends a reverse ARP request to the customer, asks the node's IP address and ATM address, and stores the request in its ATMARP table. Then, any node in LIS That wants to resolve the destination IP address will send the ATMARP request to the server. If the address ing is found, the server will return the ATMARP response. Otherwise, it returns an ATM_NAK response to indicate that no such ing exists. The server regularly clears the address ing table unless the customer responds to its periodic reverse ARP request. Once the LIS client obtains the ATM address corresponding to the IP address, it can establish a connection with the address. The Protocol for group encapsulation and address resolution is defined in RFC1483 and RFC1577 respectively.
However, because the Address Resolution Protocol defined in RFC1577 retains the host's requirements for the default router to send packets to sites outside the subnet, the shortcut VCC can only be created between nodes in the same subnet, otherwise, the source site must forward the group to the default router, even if the source and target sites are in the same ATM network. In this way, the ATM router becomes a bottleneck and the service quality cannot be achieved.
Compared with LANE, RFC1577 only supports IP addresses, rather than other network layer protocols, such as IPX and AppleTalk. In addition, CLIP does not support multicast, which is also an important drawback of RFC1577.
2. CLIP Extension
2.1. NHRP (NextHopResolutionProtocol)
In order to provide shortcuts between sites in the same ATM network and different subnets, IETF proposed the protocol named NHRP. NHRP is built on the CLIP model, but non-broadcast Multi-Channel Access Network (NBMA) instead of LIS, NBMA allows multiple devices to connect to the same network, but can be configured to different broadcast domains, and supports direct communication between hosts in different LIS. Frame Relay and X.25 are examples of NBMA networks.
NHRP replaces the ARP Server with the concept of the National Health Service (NHRP server). Each national health service contains a "Next Hop Resolution" cache table, the content is the IP address ing from all nodes related to the MNS to the ATM address. The node is configured with an ATM address containing the MNS, and its own ATM address and IP address are registered with the MNS registration package.
The protocol processing process is as follows: when a node wants to send a group through the NBMA network, that is, it needs to parse a specific ATM address, it generates and sends an NHRP request packet to the aicloud, such requests and all NHRP information are sent through the IP package. If the target site is provided by the MNS service, the MNS returns its address through the MNS response package; otherwise, the MNS searches for its route table to determine the next mns to reach the destination and forwards the request. Execute the same algorithm at the next ISP until the requested NHS ing is truly known. The destination node returns an NHRP response, which goes through the same series of ISPs in reverse order and reaches the request node, the request node can establish a direct data connection. In this way, you can establish ATMVCC over the subnet boundary so that the subnet can communicate without passing through the route.
2.2 Multicast
Two multicast methods are supported.
The first is to establish a point-to-point connection between all nodes that want to send multicast information through the multicast server. It is connected to all receiving nodes through point-to-point connections. The multicast server receives data through the point-to-point connection and resends data through the point-to-point connection. This method can be used in large networks, but multicast servers may eventually become bottlenecks.
The second method is multicast network. each node in the group establishes a point-to-multiple-point connection with other nodes. In this way, all nodes can send data to and receive data from other nodes. For a group containing N nodes, it will require N points to multi-point connections, not suitable for groups with a large number of nodes.
Both methods are used for the multicast address resolution server (MARS) recommended by Armitage ). MARS serves a cluster node. All the end systems in a cluster are configured with the ATM address of MARS. When an end system wants to send mass information to a specific group, it establishes a connection with MARS and sends out the MARS_REQUEST information. MARS returns the MARS_MULTI information, this information contains the addresses of multicast servers or group members in the group. If the Group supports multicast servers, the request node establishes a connection with the server, send data to the server, which forwards the data to nodes in the group. In the multicast network solution, the request node establishes a point-to-multiple connection with the node in the group and sends data through the connection.
　
4. MPOA
1. MPOA principles
MPOA aims to effectively transmit unicast data between subnets in the LANE environment. MPOA integrates LANE and NHRP to retain the LANE and improve the efficiency of Inter-subnet communication through the bypass router. MPOA allows physical separation of Network-layer route computing and data transmission. This is called virtual routing. Route computing is executed by the server in the vro, that is, MPS, and data transmission is performed by the customer in the edge device, that is, MPC.
At the entry point, the MPC detects the data flow sent to a router containing MPS through ELAN. When it finds a shortcut that can bypass the current route path, it uses NHRP-based protocol requests to establish shortcuts with the target node. If feasible, the MPC records the information in its entry table, creates a shortcut VCC, and sends frames through the shortcut VCC. For groups that use shortcuts, the MPC removes the data link layer (DLL) encapsulation from the group.
At the exit point, the MPC receives network data from other MPCS. For frames received through shortcuts, the MPC adds appropriate DLL encapsulation to send them to the upper layer protocol. The DLL encapsulation information is provided by MPS and stored in the egress cache.
MPS is the logical component of a vrohrp and provides network-layer forwarding information to the MPC. It contains the complete MnS defined in NHRP. MPS interacts with the local health service and routing function to answer MPOA requests from the access to the MPC, and provides DLL encapsulation information to the exit of the MPC.
The following is a brief description of the communication process between ELAN and ELAN.
ELAN Intranet Communication is from one MPOA host or LAN host to another MPOA host or LAN host of the same ELAN. These data streams use ELAN for address resolution and data transmission. ELAN communicates with each other from an MPOA host or LAN host to an MPOA host or LAN host of different Elans. Short data streams use the default path, while long data streams use shortcuts. The default path uses the ELAN and router, shortcuts use LANE and NHRP for address resolution and shortcuts. The shortcut works like this: if the source node and the target node are not in the same MPS management domain, the MPS portal translates the MPOA resolution request into an NHRP resolution request, use NHRP to forward the request to the egress MPS. When the egress MPS receives a response from the egress MPS, it generates an NHRP resolution response and sends it back to the portal MPS, after receiving the MPOA resolution response from MPS at the entry point, a shortcut can be established between the MPOA at the entry point and the exit point.
2. advantages and limitations of MPOA
MPOA separates data transmission from route computing by distributing functions to different devices, reducing the number of devices involved in route computing and the complexity of end devices. It supports layer-2 and layer-3 network interconnection in a unified manner, thus ensuring large-scale interconnection in the ATM environment. It can effectively process burst data and long-term data streams at the same time, but the complexity of MPOA is controversial.
　
V. IP exchange
The purpose of IP address switching is to obtain the most effective IP address implementation on the fast switching hardware. The advantages of non-connected IP addresses and connection-oriented ATM are complementary. IP address switching is a standard ATM switch and smart software controller connected to an ATM Switch Port, that is, an IP address Switching Controller. The IP Switch sends the initial data stream group to the standard routing module (a part of the IP Switch) for processing. When the IP Switch sees a sufficient group in the stream, it is considered to be a long-term one, A flow flag is created for an adjacent IP switch or edge device, and subsequent groups can mark the exchange at high speed to bypass the slow routing module. A special IP exchange gateway or edge device is responsible for the conversion from a non-marking group to a marking group and a grouping to an ATM data.
Each IP exchange gateway or edge device that connects an existing network device to an IP Switch establishes a virtual channel to the IP exchange controller at startup as the default forwarding channel, when an existing network device receives a group, the edge device sends the group to the IP address exchange controller through the default forwarding channel.
The IP Switching Controller executes traditional routing protocols, such as RIP, OSPF, and BGP, and forwards the packets to the next node through the default forwarding channel in normal mode, this may be another IP switch or edge device. The IP address exchange controller also performs data stream classification, which recognizes long-term data streams because such data can be optimized using the cut-through switch of the ATM hardware, and the rest of the communication still uses the default method, it is a point-to-point storage and forwarding route.
When a long-term data stream is identified, the IP exchange controller requires the previous section to mark it and use the new virtual channel. If the source edge device agrees, the data stream flows to the IP exchange controller through the new virtual channel. The next node also performs the same action. When the stream uses a special input channel and output channel independently, the IP exchange controller instructs the switch to establish an appropriate hardware port ing, bypass routing software and related processing expenses. This process continues. The first several groups of the stream allow direct connections from the source edge device to the target edge device. This design enables the IP Switch to forward groups at a rate limited by the switch engine only. The first generation IP switch supports a throughput of up to 5.3M per second. In addition, because the ATM cells do not need to be encapsulated into the IP Group of the intermediary IP switch, the throughput in the IP network is also optimized.
Ipsilon provides two protocols for IETF. The Common Exchange Management Protocol (GSMP, RFC1987) allows IP switch controllers to access switch hardware and dynamically change the switching mode: storage forwarding or cut-through. Ipsilon Traffic Management Protocol (IFMP, RFC1953) is used to exchange control information between edge devices and IP exchange controllers and associate IP streams with ATM virtual channels.
An important feature of IP address switching is that stream classification and switching are performed locally, rather than end-to-end based, which retains the non-connection nature of IP addresses, the IP switch is allowed to bypass the routing of invalid nodes without re-establishing a channel from the source host.
In addition, stream classification allows IP address exchange to support both long-term and burst data.
However, IP exchange is stream-based, and its scalability in large networks is questionable. The number of streaming in a large network may eventually exceed the number of available virtual channels.
Five companies officially claim to support Ipsilon IP exchange, including Ericsson, GeneralDatacomm, HitachiAmericaLtd., NECAmericaInc. And DECIpsilon. They try to make this technology a de facto standard-MPLS.
　
Vi. tag switching
Another option is Cisco's tag exchange. The tag switching network consists of three components: the tag edge router, the tag switch, and the tag distribution protocol.
The Edge Routers are located in the routing devices with full Layer 3 functions at the edge of the switched network. They check the incoming groups and add the appropriate labels before forwarding to the marked switched network, delete this tag when the group exits the tag switching network. As a fully functional router, the edge router can also apply value-added layer 3 services, such as security, fee recording, and QoS classification. The capability of marking an edge router does not require special hardware. It is implemented as an additional feature of Cisco software. The original vro can be upgraded with the feature of marking the edge router through software.
A tag switch is the core of a tag switch network. Tags are short and fixed-length tags, allowing the tag switch to use fast hardware technology for simple and fast Table query and group forwarding. The tag can be located in the VCI domain of an ATM Cell, the flowlabel domain of IPv6, or between Layer 2 and Layer 3 header information, which enables the tag exchange to be used on a wide range of media, including ATM connections and Ethernet.
The tag Distribution Protocol provides a method for marking switches and other tag switches or for marking edge routers to exchange tag information. Vrouters and vswitches use standard routing protocols (such as BGP and OSPF) to establish their routing databases. Adjacent tag switches and Edge Routers distribute tag values in Tag Information Library (TIB) by TAG distribution protocol.
The following describes the basic process of marking the exchange network.
(1) Mark Edge Routers and marked switches use standard routing protocols to identify routes. They can fully interoperate with non-marked switches.
(2) edge routers and switches are marked and distributed by marking the distribution protocol to the route table generated using the standard routing protocol. Edge Routers are marked to receive and label the distribution protocol information and create a forwarding database.
(3) When a marked edge router receives a group that needs to be forwarded by marking the exchange network, it analyzes its network layer header information, executes available network layer services, and selects a route for the group from its routing table, the tag switch that is marked and forwarded to the next node.
(4) When a tag switch receives a tagged group, it only performs tag-based exchange without analyzing the network layer header information.
(5) The Edge Router marked by the Group to reach the exit point, marked as stripped, and then forwarded.
In a tag exchange network, the tag distribution protocol and standard routing protocol can be combined with the target prefix marking algorithm, which can create Tag Information in TIB before the data stream passes through the network. This has two meanings. One is that all groups in the stream can be tagged for exchange, even for burst short data. In addition, it is topology-based and assigned a tag for each source/destination. In IP address exchange, a shortcut is established only after a certain number of groups have long-lived data streams. Therefore, tag switching is more effective than the stream-based mechanism to use tags, avoiding the creation process of a stream, which makes it highly scalable for public Internet service networks, in the public internet, the number of streams is huge, and the change rate is also very high.
Other vendors also have similar mechanisms, such as Cabletron SFVN, Cascade IPNavigator, DEC IPpacketswitching, FrameRelayTechnologies and ibm aris (AggregateRoute-basedIPSwitching.
VII. Conclusion
This article briefly introduces some solutions that support IP addresses on an ATM network. These solutions are based on the assumption that a traditional LAN is connected to a router through an ATM network, or, the hardware platform is an ATM network, and the application is IP-based. Other content is not described here.