所有参考文献均出自IEEE文档，和 IEEE Meeting Proceedings.
文章主要盖括了和ATM网络有关的几种IP应用。希望会对需要的人有帮助。
1. Introduction
During the second half of past decade, Asynchronous transfer mode (ATM) has reached its technological maturity and received much attention due to its high capacity, its bandwidth scalability and its ability to support multi-service traffic. (Newman, P and Minshall, G and Lyon, T 1998:117) ATM is connection-oriented, which is clearly distinct from other modern connectionless protocols, and it is able to support strict Quality-of-Service (QoS) on each connection. The widespread Internet Protocol (IP) technology has also been dramatically developed in the last several years. However, IP is a connectionless protocol and is designed to provide internetworking functions capable of running over a variety of network technologies. Moreover, IP is a network-layer protocol supporting both point-to-point and multipoint-to-multipoint (or multicast) data communication services. (Guarene, E and Fasano, P and Vercellone V 1998:74)
Due to the different features of these two technologies, wide researches have been done on the integration of these two network architectures in order to provide more efficient, capable, and reliable networking environment that meets the Quality-of-Service (QoS) requirements. Many ideas and proposals are discussed widely around the possible integration of IP and ATM. These proposals can be recapitulated as following several issues. Firstly, the mismatch of IP and ATM leads to complexity and inefficiency in attempting to transfer connectionless IP datagram over the connection-oriented ATM network. In order to make these two technologies cooperate without obstruction, several approaches must be established to solve this problem. Secondly, with IP technology, datagram is routed hop by hop towards its destination without any connection establishment between two parties, according to the destination’s unique IP address contained in the IP header. On the other hand, ATM uses its own address resolution. Therefore, in order to binding these two address resolution, a new address resolution protocol is absolutely required to map IP address onto ATM address as well as ATM address onto IP address. Thirdly, multicasting is a powerful capability of IP because many multimedia applications involve data delivery from one source to multiple destinations. However, ATM itself does not support multicast services, which results in the requirement of the capability of handling multicast when such applications are applied over ATM. Furthermore, real-time application like video and audio conferencing require the network to provide both QoS guaranteed services and best effort service; therefore a new protocol must be developed to fulfill the requirement. (Cocca, R and Salsano, S and Listanti, M 1999:98)
Consequently, a series of approaches are developed to support the above capabilities in the integration of IP and ATM, including “classical IP over ATM”, Next Hop Resolution Protocol (NHRP), Cell Switch Router (CSR), Multiprotocol Label Switching (MPLS) and Resource Reservation Protocol (RSVP). (White, P 1998:79) (Guarene, E and Fasano, P and Vercellone V 1998:74)
This report briefly introduces the classical IP over ATM in section 2. Section 3 presents the issues relating to the IP Routing and ATM Switching with various protocols. Section 4 presents the relating issues in multicasting over ATM network. Section 5 discusses the internet integrated services over ATM. Finally, section 6 will discuss IPv6 and ATM.
2. Classical IP over ATM
The first solution to combine IP and ATM was to follow an overlay model which considers ATM as a data link layer, according to which, the “classical IP over ATM” was developed by the IETF (Internet Engineering Task Force). (Guarene, E and Fasano, P and Vercellone V 1998:74) This model does not change IP in any way, and use ATM as an underlying technology, that is, separating the ATM network into a number of logical IP subnets (LISs) interconnected by IP nodes (such as hosts or routers). With this model, direct ATM connection between IP hosts in different LISs are prohibited even if the underlying ATM topology supports them. Instead, an ATM Virtual Channels (VC) is introduced to set up the connections which allow the IP packets or ATM cells to be transferred within a single LIS. Nevertheless, once the transmission reaches the router at the LIS boundary, the ATM cells must be reassembled into IP packets, each of which is then subjected to an IP forwarding decision and re-segmented into ATM cells to be sent along the next LIS VC, in which case, end-to-end communications requires n + 1 VCs where n is the number of routers to be crossed. (White, P 1998:79) ( Guarene, E and Fasano, P and Vercellone V 1998:74)
The mappings between ATM address and IP address within a single LIS is provided by an ATM Address Resolution Protocol (ARP) server and each LIS supports a single ATM ARP server. (Park, C and Lee, J and Kim, J and Kim, H 1997:378) The configuration and operation of an ATM ARP server in a LIS is similar to an ARP server in a traditional IP subnet. When a node comes up within a LIS, a connection to the ATM ARP server is established, using configured address. Subsequently, an ARP_REQUEST packet is sent from an end system to the ATM ARP server to resolve the ATM address of an IP address. If an address mapping that matches the IP address is found, the corresponding ATM address will be returned to the inquiring end system in an ARP_REPLY packet. Otherwise, an ARP_NAK packet is returned. Once a LIS client has obtained the ATM address corresponding to a particular IP address, it then setups a connection to the address using signaling procedures. (Park, C and Lee, J and Kim, J and Kim, H 1997:378)
However, such address resolution mechanism makes it difficult to communicate between two nodes on two different LISs on the same Logical Address Group (LAG) due to traversing each router on the intermediate hops on the path between the source and destination. (Park, C and Lee, J and Kim, J and Kim, H 1997:378) Consequently, the optimization of addressing is required to improve the performance and scalability of the integrated network, including both “improving the router throughput by merging IP routing with ATM switching” and “redesign new overall network architecture to cope with future communication requirements”. (Guarene, E and Fasano, P and Vercellone V 1998:74)
3. IP Routing and ATM Switching
3.1 Next Hop Resolution Protocol (NHRP)
As mentioned before, with Classical IP over ATM, inter-LIS communication has to go through routers, which is not an optimal solution when both parties involved are attached to the same ATM network. To utilize the potential benefits of ATM which are lost with the classical models, the Next Hop Resolution Protocol (NHRP), used to establish unicast ATM VC’s that bypass IP routers, is defined within the IETF. (Guarene, E and Fasano, P and Vercellone V 1998:74)
NHRP is an extension of ATMARP that for an end system to resolve the IP address of another end system in a foreign LIS into its corresponding ATM address. (White, P 1998:79) NHRP is based on a client-server approach and consists of two types of entities, Next Hop servers (NHSs) and NHRP client (NHC), and protocols between them. Each LIS contains at least one NHS and each end system is a NHC. When an end system needs to resolve an IP address, it sends a request to NHS in charge of his LIS. If the IP address to be queried belongs to the LISs it is serving, NHS expects to find an entry that matches the IP address and replies with the corresponding ATM address. Once the source receives the response, it can then open a direct cut-through VC to the destination using standard ATM signaling/routing protocol. (White, P 1998:79) Otherwise a negative reply will be returned.
Eugenio, Paolo and Vinicio (Guarene, E and Fasano, P and Vercellone V 1998:74) introduces an improved usage of NHRP that the classical hop-by-hop routing and shortcut routing (e.g., direct SVCs) can be chose by NHRP client according to the user application. Although a dedicated SVC can be suitable for long data transferring and QoS guaranteed services (e.g., video and audio streams), it is too costly, in terms of network resources and connection set-up latency, due to the short data transferring, which do not care about QoS. (e.g., remote login)
Unquestionably, NHRP overcomes some of the weaknesses of the classical IP over ATM, it still have its own limitations. One of these is that the very large ATM networks with NHRP technology become complex and unmanageable. Another critical limitation is the inability to directly support multicast. This will be discussed in section 4.
3.2 Cell Switch Router (CSR)
Cell Switch Router (CSR), a class of new network elements, was introduced by Toshiba and is receiving increasing attention in the worldwide research and industrial community. This new architecture is an internetworking device that integrates IP routing and ATM switching allowing coexistence of hop-by-hop IP forwarding with direct VC cut-through modes of service in order to improve IP routers performance and cost characteristics. (White, P 1998:79) ( Guarene, E and Fasano, P and Vercellone V 1998:74)
Similar to the process of the conventional router, the traffic received from the ATM interfaces is sent to the routing modules, and datagram are processed and forwarded toward their destination. Simultaneity, however, IP packets are monitored by the CSRs, according to some policy, deciding if any particular sequence of packets (i.e., flow) may take advantage of straight ATM switching, in which case, CSRs set up new VCs dedicated to the specific flow and concatenate incoming and outgoing VCs in the internal ATM fabric performing a cut-through path. Subsequently, ATM cells are directly forwarded bypass routers at the boundaries of each LIS. (Guarene, E and Fasano, P and Vercellone V 1998:74)
Compared to NHRP and traditional ATM signaling, this cut-through service is different due to the independent decision on whether or not to implement local cut-through made by each switch router. (White, P 1998:79) “ATM connections are managed integrating IP and ATM control functions.” (Guarene, E and Fasano, P and Vercellone V 1998:74)
3.3 Multi-Protocol Label Switching (MPLS)
The IETF (Internet Engineering Task Force) introduced a new internet architecture aiming at an optimal integration of IP and ATM functionalities which is called Multi-Protocol Label Switching (MPLS). (Guarene, E and Fasano, P and Vercellone V 1998:74) MPLS supports all network services in the future Internet including multicast and QoS, by “replacing the standard destination-based hop-by-hop forwarding paradigm with a label-swapping forwarding paradigm.” (Visvanathan, A and Feldman, N and Wang Z and Callon R 1998:165)
A label is appended to each packet in MPLS on the underlying ATM network. To enhance the IP networking performance, this label allows packets to be routed at high speed to the appropriate egress interface on a router. It also associates packets with a continuous path through the network which is called the label switched path (LSP) that enhances the management of the network. (Bocci, M and Guillet, J 2003:139)
Matthew and Jim (Bocci, M and Guillet, J 2003:139), however, point out that similar to many new technologies, “the excitement of things new can detract from an appropriate application analysis driven by business requirements.” Since each service providers’ business objectives and the services they offer vary, the appropriateness of ATM-MPLS network interworking can only be discussed within each service providers’ unique context. MPLS is a powerful new tool for building networks, but as with all new technologies, it should be used wisely.
4. IP Multicast over ATM
Eugenio, Paolo and Vinicio (Guarene, E and Fasano, P and Vercellone V 1998:74) defines IP multicast as a multipoint-to-multipoint service that allows dynamic management of multicast groups (hosts can join and leave a multicast group at any time), and provides packet replication and distribution tree creation functions. As mentioned before, “Classical IP over ATM” and NHRP only support IP unicast over ATM. To support IP multicast, two issues need to be solved. First, we need an address resolution protocol to translate a multicast IP address into a list of ATM addresses, and this is solved by Multicast Address Resolution Server (MARS). Second, we need to specify how multicast data is transferred among the involved parties, and VC mesh and Multicast Server (MCS) are two possible solutions.
A Multicast Address Resolution Server (MARS) is introduced into each LIS to perform the multicast address resolution. It answers the queries for multicast addresses from the end systems in the same way as ATMARP server answers queries for unicast addresses. An end system joins or leaves a particular multicast group by sending Internet Group Multicast Protocol (IGMP) packets to the MARS.
When a multicast IP address is resolved into a list of end points, the data needs to be forwarded among the group members, from the sender to the receivers. One way to do this is to let each group member set up a point-to-multipoint connection with all other group members and this approach is called VC Mesh. The other way is to introduce a Multicast Server (MCS) into each LIS that supports multicast. When an end system queries for a multicast address, MARS will reply to it with the ATM address of the MCS. The end system then sends the multicast packets to MCS. The MCS will build a point-to-multipoint connection or multiple point-to-point connection to the group members to forward the packet received from the end system to all the members of the group specified in the address field of the multicast packet.
VC mesh and MCS each its pros and cons. With MCS, if the membership of a multicast group changes, it only needs to modify the point-to-multipoint VC to the group members while with VC mesh, all connections in this "mesh" have to be modified. However, MCS needs to reassemble the packets sent from the source and resend them to the group members. With VC mesh the reassembling is not needed so the latency is minimized.
Eugenio, Paolo and Vinicio (Guarene, E and Fasano, P and Vercellone V 1998:74) introduce two different implementations of IP multicast over ATM. One is MBone over ATM by which a user can access the network by means of an ATM PVC between its private router and a backbone router using a set of multicast routers interconnected through a mesh of ATM PVCs. Another one is High-Performance multicasting, in which, the process that the multicast routers perform multicast datagram replication has been moved from IP level to ATM layer, due to the limitation of IP routers especially increasing the bit rate of multicast flows and the number of copies to be replicated.
5. Internet Integrated Services over ATM
The Internet is based on the best effort service models which provide no guarantee on the correct and timely delivery of data packets, due to which, Internet Integrated Services (IIS) architecture is introduced by Internet community to provide support to real-time applications with guaranteed quality. In IIS architecture, an important signaling protocol called Resource Reservation Protocol (RSVP) has been defined. (Cocca, R and Salsano, S and Listanti, M 1999:98)
A non QoS-oriented NHRP has been introduced in section 3. Since the growth importance of QoS required by industries, NHRP has to be improved “using NHRP and the ATM shortcuts whenever possible, including for transport of RSVP messages”. (Cocca, R and Salsano, S and Listanti, M 1999:98) Different VCs deliver different packets, including “best effort data packets for different destinations, packets of QoS flows, and RSVP control messages.” Another approach, however, is to separate the services that ATM shortcuts only used for QoS flows while best effort packets should follow the hop-by-hop path.
6. IPv6 over ATM Networks
Internet Protocol version 6 (IPv6) is a new version of Internet Protocol version 4 (IPv4). IPv6 has fixed header providing faster IP forwarding hardware implementations. Also, the flow labels differentiate the connections between the same source and destination. Furthermore, IPv6 provides enlarged capabilities allowing more addresses are used, and also integrated security and authentication methods. (Loukola, M.V. and Slytta, J.O. 1998:548)
There are many issues concerning how ATM supports this new protocol, for example, capabilities supporting QoS. One of the approaches is to map the priority field in IPv6 to the ATM cell in order to enable IPv6 over ATM support QoS guaranteed services. This approach improves the performance of high priority traffic and it does not reduce the total cell loss but “protects the high priority traffic from cell loss while allowing the performance of the low priority traffic to degrade as little as possible.” In terms of label switching, “IPv6 destination-site label switching” provides high performance of switching “a high percentage of packets in the core of the network while maintaining low VC usage.” (Boustead, P and Barnett, S and Chicharo, J, and Anido, G 1998:561)
7. Summary and Conclusion
Since the multimedia communication becomes more import, together with the growth number of IP-based applications, ATM network architecture must be improved with some extra capacities in order to fulfill the requirements of today’s industry. During the second half of past decade, ATM and IP have been dramatically developed. However, these two architectures have different features and capabilities. ATM is connection-oriented while IP is connectionless. Also, ATM supports QoS guaranteed services on each connection but IP only supports best effort services. Furthermore, IP supports multicasting, a multipoint-to-multipoint service, but ATM only supports unicast that is point-to-point service. Due to the advantages and disadvantages of ATM and IP, it is urgently necessary to integrate these two networking architectures to provide more reliable services. As a consequence, many approaches have been developed or drafted, including “classical IP over ATM”, Next Hop Resolution Protocol (NHRP), Cell Switch Router (CSR), Multiprotocol Label Switching (MPLS) and Resource Reservation Protocol (RSVP).
Of course, there are too many protocols and services classes are being develop except the protocols mentioned above. This report, however, cannot cover all the issues relating to the use of IP over ATM. The report first explain the “classical IP over ATM” and then focuses on 3 main issues relating to IP over ATM, they are, IP routing and ATM switching, IP multicast over ATM and Internet Integrated Services. The use of IPv6 over ATM is also mentioned at the end of the report.
First, the “classical IP over ATM” has been briefly described in section two. This model negates the potential benefits of ATM in exchange fro maintaining the classical IP paradigm in order to facilitate easy migration to ATM. The model splits the ATM network into a number of logical IP subnet (LIS) interconnecting with IP nodes such as IP routers. In this model, however, the IP layer is simply mapped onto the ATM layer, which leads to duplication of much functionality supported by both two architectures. For example, IP packet forwarding is overlapped with ATM cell switch. Another disadvantage is interconnecting with many IP routers requires the high performance of IP routers because the low performance of the IP router has become the bottleneck of the network. Therefore, NHRP is discussed thereafter.
NHRP is developed to improve the ATM address resolution by establishing the end-to-end VCs communications bypass the IP nodes, especially referring IP routers. This dramatically enhances the functionality of integrating IP and ATM and avoids suffering from the limitation of IP router bottleneck. NHRP, however, only support unicast, because of which, this address solution protocol does not fulfill the requirements of the industry.
The report also describes a new device called Cell Switch Router (CSR). CSR introduced by Toshiba, is a new internetworking device that integrates IP routing and ATM switching allowing coexistence of hop-by-hop IP forwarding with direct VC cut-through modes of service in order to improve IP routers performance and cost characteristics.
The Multiple-Protocol Label Switch (MPLS) is a new network architecture that supports all network services such as multimedia communication, multicasting and QoS guaranteed services. It replaces the traditional hop-by-hop IP forwarding approach with a label-swapping forwarding approach. MPLS is a powerful architecture to build networks, however, it still need to be developed. As time goes on, MPLS will reach its technological maturity.
IP multicasting is an important functionality for the ATM network to improve the performance. To merge IP multicasting functionality into ATM networking, Multicast Address Resolution Server (MARS) and VC mesh and Multicast Server (MCS) are introduced. MARS translate a multicast IP address into a list of ATM addresses, VC mesh and MCS specify how multicast data is transferred among the involved parties, and are two possible solutions. In the report, we described the basic working theory of multicasting, and then introduced two implementations which are MBone over ATM and High-Performance Multicasting. These implementations dramatically enhance the capability of IP over ATM architecture.
The report also discussed the Internet Integrated Services. As mentioned before, the number of multimedia applications has grown very fast in the last few years and will keep growing in the future. Therefore, a reliable networking architecture supporting QoS guaranteed services and real-time application urgently needs to be developed. As a result, RSVP, an important signaling protocol has been described in the report focusing on the integration of NHRP and RSVP.
IPv6, the next generation of Internet Protocol, provides more capabilities improving the current IPv4. The integration of IP and ATM, as a result, will take advantage from IPv6 and provide more fast forwarding paradigm and QoS support.
To sum up, the integration of IP and ATM provide more powerful network to support various services. Due to the mismatch of these two protocols, there are many problems to be solved. As time goes on, new technologies will be developed which probably will replace the current network technologies in order to provide more reliable networks.
Bibliography
Bocci, M and Guillet, J (2003) ‘ATM in MPLS-Based Converged Core Data Networks’, IEEE Communications Magazine 41:1 139-145 (January)
Boustead, P and Barnett, S and Chicharo, J, and Anido, G (1998) ‘Label Switching and IP Version 6’, Computer Communications and Networks 1998 Proceedings, Sixth International Conference 561-565
Cocca, R and Salsano, S and Listanti, M (1999) ‘Internet Integrated Service over ATM: A solution for Shortcut QoS Virtual Channels’, IEEE Communications Magazine 36:12 98-103 (April)
Guarene, E and Fasano, P and Vercellone V (1998) ‘IP and ATM Integration Perspectives’, IEEE Communications Magazine 36:1 74-80 (January)
Loukola, M.V. and Slytta, J.O. (1998) ‘New Possibilities Offered by IPv6’, Computer Communications and Networks 1998 Proceedings, Sixth International Conference 548-552
Newman, P and Minshall, G and Lyon, T (1998) ‘IP Switching – ATM under IP’, IEEE Transactions on Networking 6:2 117-129 (April)
Park, C and Lee, J and Kim, J and Kim, H (1997) ‘IP Routing on Logical Address Goup’, Computer Communications and Networks 1997 Proceedings, Sixth International Conference 378-382
Visvanathan, A and Feldman, N and Wang Z and Callon R (1998) ‘Evolution of Multiprotocol Label Switching’, IEEE Communications Magazine 36:5 165-173 (May)
White, P (1998) ‘ATM Switching and IP Routing Integration: The Next Stage in Internet Evolution’, IEEE Communications Magazine 36:4 79-83 (April)
