Comparison of IP-over-SONET and IP-over-ATM Technologies

Source: Internet
Author: User

In recent years, the rapid development of Internet has put forward higher requirements for Wan bandwidth. Internet service providers (ISPs) have been looking for faster trunk router interconnection methods. Recently, the Inernet solution has led to the deployment of multiple IP-over-ATM technologies by many ISPs, such as LAN simulation and general IP over ATM technologies. Because the underlying infrastructure provided by operators is the synchronous optical network SONET deployed on the wide area optical fiber link or the Synchronous Digital layered SDH, more and more people are paying attention to directly running IP addresses on SONET, rather than using ATM networks, to improve bandwidth efficiency. However, there has been a lot of controversy in the industry regarding which technology provides the best solution.
I. IP-over-ATM
The Internet Engineering Task Team IETF defines IP addresses as a public "adhesive" that connects multiple heterogeneous networks into a large interconnected network ". Different networks can adopt different physical layer, link layer, and network layer technologies. However, if the IP layer runs on the top of each network layer, You can seamlessly connect multiple networks. The Internet Engineering Task Team defines various network technologies used to run IP addresses, including broadcast LAN technologies such as Ethernet), circuit switching WAN technologies such as X.25) and group-switched WAN technologies such as the exchange-type multi-region data service SMDS ). In early 1990s, the number of deployed ATMs increased significantly. At this point, IETF began to define the operation of IP addresses on the ATM. The ATM Forum also began to define different layer 2 and Layer 3 protocols for operations on the ATM, mainly from the LAN perspective. The results of these work are summarized below.
1. Generally, IP over ATM (CIP ). Generally, IP Over atm cip allows existing IP users to switch to ATM as the underlying data transmission technology, while still using existing applications designed for traditional IP systems. An ATM network is divided into multiple logical IP address subnets (LIS). These logical IP address subnets communicate with each other through routers. Because ATM does not have the Native Broadcast function, the traditional broadcast address decomposition protocol ARP is replaced by the ATMARP protocol based on the client-server. CIP adopts the default 8-byte encapsulation, uses Logical Link Control/subnet Access Control LLC/SNAP, and transmits IP addresses and ATMARP groups at the ATM. These groups are directly mapped to ATM cells using AAL5. These cells are transmitted in the activated virtual connection PVC or the switched virtual connection SVC. To establish SVC, a signaling protocol such as ATM Forum UNI 3.1/4.0 or ITU-T Q.2931 is required.
2. LAN simulation. According to the definition given by the ATM Forum, LAN simulation LANE is a service provided on the ATM network. It simulates the existing 802.3 Ethernet LAN and 802.5 ring LAN services. By using LAN simulation, existing LAN applications can communicate on an ATM network, just as using interfaces similar to MAC drives to connect to a traditional LAN. The services provided include connectionless single-point broadcast and multi-point broadcast data transmission services. LANE runs on the MAC layer and can be used with any layer-3 protocol. In contrast, generally, IP over ATM can only run with IP addresses. Simulation lan elan) includes a set of LAN simulation clients on the ATM endpoint and a LAN simulation service.
LANE provides multiple service components for various features, such as address decomposition, configuration, and broadcast. These service components can be installed on an ATM Endpoint or vswitch and implemented in a centralized or distributed manner. An ELAN is similar to a lan cidr Block and can communicate with other lan cidr blocks through a bridge or router. To transmit LANE control and data packets through ATM, it uses a 2-byte LANE encapsulation. The new LANE specification also allows the use of LLC/SNAP encapsulation. The LANE group uses the AAL5 transmitted in SVC to map directly to the ATM Cell.
3. MultiProtocol over atm mpoa. The ATM Forum defines MultiProtocol over atm mpoa to overcome the main disadvantages of LAN simulation and CIP, that is, these protocols require hosts on ELAN or LIS of different subnets to communicate through intermediate routers, this greatly reduces the throughput of the group, because each router must reorganize the cells of the layer-3 group during routing, and divide the group into tokens during forwarding. MPOA allows clients in different subnets to establish direct VCC or a shortcut between each other, and then directly forward the group on the third layer without any intermediate restructuring or splitting process. In the subnet, MPOA uses LANE. MPOA clients can be mounted on the host connected to the ATM or as part of the edge device that connects the ATM subnet to the non-atm lan cidr block. These edge devices can be connected to the VCC forwarding group on the L2 or L3. Therefore, the MPOA client monitors layer-3 group traffic. If a continuous group flow is found for a specific sink, the customer will ask the MPOA server to provide the ATM address, to establish a shortcut VCC to the sink. The MPOA server interacts with the routing server of the ATM subnet to obtain necessary information. You may need to use the IETF-defined next-site decomposition protocol NHRP along the route path) to spread the decomposition request to the queried destination until you obtain the ATM exit point closest to the requested destination. MPOA is used in Distributed Virtual routers. Connecting an ATM subnet to an edge device in a traditional lan cidr block is similar to an interface card used by a vro. The entire ATM network connecting edge devices is the vro forwarding baseboard. The Group forwarding function is separated from the routing computing function, which is completed by the Routing Server. Group Forwarding is separated from route computing, which improves the efficiency and throughput compared with traditional routers. The group uses the LANE or LLC/SNAP encapsulation for transmission. These groups are directly mapped to ATM cells using the AAL5 transmitted in SVC.
4. multi-protocol standard switch MPLS. At present, the industry has concentrated a lot of work on developing a solution that integrates on ATM switches rather than overlapping IP Routers. Many vendors have proposed a solution to integrate IP addresses and ATMs. IETF has already specified an integrated mechanism called Multi-Protocol Label Switching MPLS ). In MPLS mode, each vro is also a vswitch. In addition to the Third-layer packet header, the group assigned to the shortcut path also has a fixed length mark. Through these tags, you can use a fixed-width search table, while a traditional router uses the longest prefix matching method), very quickly forward groups. MPLS will define its own label Allocation Protocol LDP, which will closely interact with the layer-3 routing protocol and establish the desired shortcut path. In addition, MPLS allows you to establish shortcuts based on a variety of standards, such as sink IP addresses, service categories, and service policies, so that you can flexibly design network engineering. MPLS is not limited to ATM. On the contrary, MPLS is designed for any link layer technology. As long as this technology supports identifying fixed-length labels of shortcuts, MPLS can all run on this technology.
Ii. IP/PPP/SONET
Based on the IETF development work and announced vendor plans, the current industry direction is to adopt the Point-to-Point Protocol (PPP) as a mechanism for transmitting IP address digital reports in the SONET server load. IP/PPP/SONET maps IP packets to SONET payload at STS-1 51.84 Mbit/s, STS-3c 155.52 Mbit/s, and STS-12c 622.08 Mbit/s rates.
The IP/PPP/SONET ing shows how to use PPP, and the routing signal is marked as part of the byte SONET path overhead) is set to 207. The data unit of the PPP protocol is the same as that of the SONET payload. It provides multiplexing functions through the "protocol" field and allows the transmission of data units of other protocols, such as IPX. PPP also includes control protocol link control protocol and network control protocol to start and configure link layer protocol and network layer protocol. PPP is simple, stable, and widely used. Ietf rfc that can use this method includes: RFC 1661, W. simpson, "Point-to-Point Protocol PPP"; RFC 1662, W. simpson, "PPP is used in frame sharding similar to HDLC"; RFC 1619, W. simpson, "PPP over SONET/SDH"; RFC 1700, J. renault, J. postel "quantity allocated ".
Iii. Comparison between IP-over-SONET and IP-over-ATM
Compared with IP-over-ATM, IP-over-SONET operations have many differences. Some important issues are summarized as follows.
1. Protocol overhead. So far, the maximum reason for ISP to consider deploying IP-over-SONET instead of IP-over-ATM is that the ATM cell header contains 5 bytes per 53 bytes, it is sometimes called a cell tax. AAL5 padding, 8-byte ending) and LLC/SNAP encapsulate 8 bytes) also increase the overhead.
When running on an ATM, the IP address only achieves the available line rate of about 80%, while when running on a SONET, it can achieve the line rate of 95%. When expensive wide-area links or other links restricted by bandwidth are used for trunk router interconnection, the increased IP-over-SONET capacity is extremely attractive. It is not a big problem for scenarios with sufficient bandwidth, such as local area networks and bandwidth efficiency.
2. bandwidth management. ATM provides a full range of functions to manage the bandwidth allocated to the various information flows through a link (VCC. It allocates flexible bandwidth for these VCC instances based on the required service quality. Due to its cell exchange characteristics, ATM allows multiple information flows to share the same link at the same time, and ensures that a certain amount of bandwidth is allocated to each information flow. However, PPP does not provide any bandwidth management functions. It provides a simple point-to-point link, and the IP layer must schedule its grouped transmission to ensure that each information stream gets a fair share of the link bandwidth. Problems may occur on slow links because large groups of low-priority information flows may block the transmission of other groups with high priority. For example, large groups in a file transfer stream with a lower priority may have a much lower latency but are more time-sensitive. IP-over-SONET provides a bandwidth efficiency advantage for latency-sensitive real-time applications on a bandwidth-restricted link, which may offset this advantage to some extent.
3. service quality. QoS is related to parameters such as end-to-end group latency, jitter, packet loss, and throughput. ATM provides a rich set of QoS parameters that can be determined by each VCC negotiation. The smart queuing and Scheduling Mechanisms in the vswitch ensure that the negotiated QoS can be provided. ATM provides various service levels to meet different application requirements. For example, applications with very special QoS requirements can use the constant bit rate CBR) or variable bit rate VBR) service. Applications that require elasticity can use available bit rate (ABR) or unspecified bit rate (UBR) services. These Native ATM functions allow simple QoS provision on the IP layer. On the IP layer, each information stream with specific QoS requirements can be mapped to its own VCC with specific QoS. For example, a speech stream can be mapped to a real-time CBR or VBR connection, while a file transfer can be mapped to An ABR connection. PPP runs on a single point-to-point link and does not provide any QoS functions. As mentioned above, the IP layer must intelligently manage its group transmission to ensure proper QoS for information flows. Although ATM provides a rich set of QoS parameters, QoS-based services are limited to the ATM paths connecting two routers. To provide end-to-end QoS to IP groups, the router must also provide smart queuing and Scheduling Mechanisms. In this sense, when the IP network overlaps at the top of the ATM network, the router regards the ATM connection as a point-to-point link, which is similar to PPP, although the actual communication may occur on a network consisting of an ATM switch.
4. Address and route. ATM is defined as a complete network layer, which provides a wide range of functions for terminal system addressing and connection routing. An ATM network can span a large geographical area and provides a general interconnection mechanism between routers, regardless of the locations where these routers are located. In contrast, PPP runs only on direct point-to-point links without addressing or routing. To create a master network, you must activate a point-to-point link between the master routers. You must activate multiple links to implement fault tolerance. In some cases, you may need to configure a full mesh to minimize the number of sites required to span the trunk. The full-mesh structure is not only costly, but may not be feasible because there is limited access to the pure SONET link in the wide area. When used in combination with SVC, the ATM achieves any router connection between the routers, without the need to configure a full mesh structure. Even if some links in the ATM network fail, the dynamic SVC routing function can still find a roundabout route, and always ensure the connection between any two routers. The most useful function of ATM is that the operator can easily establish a connection with other routers on an ATM interface. In the Backbone Router network, most routers need to communicate with each other, which means that they will eventually need a full mesh connection, regardless of the point-to-point link or SVC. However, ATM can also achieve more flexible network engineering design capabilities, because it can route SVC on different links and use the same access link, connect a vro to multiple sinks. The ATM traffic control protocol uses a variety of functions, such as call acceptance control CAC, communication shaping, and user parameter control UPC or policy formulation, to ensure that the information flow remains within the boundaries of the negotiated communication contract. More than one communication is marked, and these groups can be discarded when the network is overloaded. Therefore, the end user can understand the hidden information related to Congestion Based on the marked or dropped groups. The cell-level discarding of ATM and TCP's group-level traffic control and interaction capabilities are poor. To eliminate this problem as much as possible, the industry has developed a variety of mounting technologies for ATM, for example, partial or early packet loss, the EPD can be used to identify the boundary of the AAL frame of the group and discard the entire frame in case of overload. Recently, the ATM Forum defined the API service. It indicates the allowed rate at which the ATM endpoint can send communications to the network and provides feedback for traffic control. This rate can change with the network load, allowing users to obtain available bandwidth without overloading the network. In ideal conditions, the ABR will eliminate the loss of cells in the network and push the congestion to the boundaries of the ATM network. This requires the vro to buffer more groups. PPP does not provide a traffic control mechanism, so TCP traffic control runs directly on the PPP link. As mentioned above, no matter whether a vro is connected through an ATM or directly connected through SONET, The vro can view the bandwidth between the vrouters.) an appropriate buffer mechanism must be used to ensure reasonable throughput.
5. multi-protocol encapsulation. ATM provides two mechanisms for multiple protocols to share the same link. The first mechanism is VCC multiplexing, which distributes each protocol to each VCC. The ATM layer multiplexing and anti-multiplexing VCC. Therefore, you do not need to add any other encapsulation headers to distinguish different protocols. The second mechanism is LLC multiplexing, which allows multiple protocols to share the same VCC. It adds an 8-byte packet header to each group to identify which Protocol it belongs. This type of multiplexing can be used when the number of available VCC instances is limited and VCC instances are shared among various protocols. PPP provides a multi-protocol encapsulation similar to LLC multiplexing in an ATM. It uses the 1-byte or 2-byte protocol Identifier Field as the Encapsulation Header. In the most important aspect, the multi-protocol encapsulation functions of PPP and ATM are the same.
6. Fault tolerance. By using a dynamic routing protocol called the dedicated network node interface PNNI protocol), in the case of a faulty link and a route connection around the switch, ATM provides fault recovery capabilities. Currently, PNNI only provides roundabout routing during initial connection establishment. PPP does not have any fault tolerance feature because it runs on a single link. However, the underlying SONET layer provides built-in protection. When the operating halo fails, you can switch the switch to the loose halo. When running on SONET, you can also apply this function on the ATM.
Iv. Deployment Scheme
The above sections compare the relative standards of IP-over-ATM and IP-over-SONET. In essence, when IP-over-SONET is used, the router is connected through a fast point-to-point link. When IP-over-ATM is used, A router is connected through a network composed of multiple links. These links carry multiplexing connections, and each connection is related to flexible bandwidth and negotiated service quality. The difference between the two technical solutions can be attributed to the comparison of speed and flexibility. Depending on which factor is more important in a specific application, we can implement the following deployment scheme.
1. ISP trunk. ISP trunk generally requires high-speed Interconnection of the trunk router to maximize the group throughput. Therefore, ISP and its suppliers are very interested in running IP-over-SONET to interconnect the trunk router. However, IP-over-SONET lacks bandwidth management, service quality, and flexible network engineering design capabilities, thus reducing the interest of ISPs and their suppliers. In addition, high-speed SONET links can transmit packets at extremely high rates, which may be higher than the rate that most routers can process. IP-over-SONET has many advantages in connecting high-speed trunk routers on a wide-area link that does not require high service quality or has bandwidth restrictions.
2. Enterprise Intranet. Enterprise intranets across a wide range of regions face the same problems as ISP trunk. From the perspective of cost, IP-over-SONET may have many advantages. However, these advantages must be weighed against the costs required to acquire necessary equipment and services. IP-over-SONET is still an emerging market, and the device costs may be relatively high. In contrast, IP-over-ATM is quickly becoming a commodity, and increasing competition is driving down device costs. Similarly, carriers can sell ATM links at a lower price than SONET links because they provide flexible bandwidth management functions. In addition, deploying an ATM allows you to easily share the wide-area bandwidth for non-IP applications, such as SNA, IPX, Appletalk, and DECnet.
3. Park trunk. Because most of the park's trunk uses relatively low-cost physical interfaces, such as multi-mode optical fiber 100-155 Mbps, shielded twisted pair wires, or unshielded twisted pair cat5e UTP Cat-5 copper wire TAXI on 155 Mbps, therefore, these networks are less likely to deploy SONET. Even if SONET is deployed, IP-over-SONET bandwidth efficiency may not compete with the flexibility of IP-over-ATM because the bandwidth may be sufficient.
4. carrier network. Carriers are deploying SONET in their networks. They are also likely to deploy ATM over SONET to provide flexible bandwidth management functions for paying customers and ensure service quality. As a result, they can provide IP-over-ATM services more easily than simply activating IP addresses on SONET.
At present, there are two obvious development trends in the network field. The first development trend is that IP is rapidly becoming the preferred Network Layer Technology for group network construction and is promoting the development of the global Internet; the second development trend is that operators are deploying SONET extensively. SONET may become the physical facility for connecting vrouters on the wide area network. IP-over-SONET is more attractive when the original speed is critical. When flexible bandwidth management, service quality, and network engineering are critical, IP-over-ATM provides a better solution. Cost-related issues will prevail over all these issues. These costs include Service Activation costs and service maintenance costs. As a complex technology, ATM naturally leads to higher network operation and maintenance costs. According to these characteristics, ISP and some enterprise intranets are most likely to deploy IP-over-SONET in restricted environments. In contrast, the carrier and the park trunk will prefer IP-over-ATM. Emerging technologies may make IP-over-ATM more attractive. One of the major concerns is IETF's MPLS development. MPLS greatly improves the flexibility of network engineering and bandwidth management in the use of ATM. In addition, it tightly integrates IP addresses and ATM routes to eliminate the overlapping model and inefficiency of the current IP-over-ATM solution. With the introduction of the current IP-over-ATM technology, such as LAN simulation and MPOA, and the emergence of new technologies that run IP-over-ATM, the Development Trend of IP-over-ATM will become more obvious.

Contact Us

The content source of this page is from Internet, which doesn't represent Alibaba Cloud's opinion; products and services mentioned on that page don't have any relationship with Alibaba Cloud. If the content of the page makes you feel confusing, please write us an email, we will handle the problem within 5 days after receiving your email.

If you find any instances of plagiarism from the community, please send an email to: info-contact@alibabacloud.com and provide relevant evidence. A staff member will contact you within 5 working days.

A Free Trial That Lets You Build Big!

Start building with 50+ products and up to 12 months usage for Elastic Compute Service

  • Sales Support

    1 on 1 presale consultation

  • After-Sales Support

    24/7 Technical Support 6 Free Tickets per Quarter Faster Response

  • Alibaba Cloud offers highly flexible support services tailored to meet your exact needs.