ATM switch call processing architecture

1. Overview
The ATM signal system evolved from the ISDN signal system to adapt to the services provided by the broadband network. It is a new technology used in recent years. The combination of ATM circuit switching and group switching is a connection-oriented network technology, there are corresponding signaling routing rules and address structures, fast speed, large capacity, multiple businesses, strong support capabilities, good scalability, and high-reliability QOS Assurance. ATM is a connection-oriented technology that relies on signal establishment and control connections. In an ATM switch, the main components used to process connections include signal, call control, and routing, A signal is a language used to establish management and terminate connections. It allows two exchange systems of the same level to establish a call through transmission. The main function of call control is to coordinate connection management in the exchange system and execute local service management policies. The main function of routing is to find a connection channel for two or more exchange systems in the network, the selected road must meet the end-to-end service requirements.
ATM enables enterprises to adopt a single network technology between the LAN and WAN to provide a truly perfect LAN/WAN integration. ATM not only guarantees the user level, but also ensures the user's latency and packet loss rate. The QOS advantage of ATM makes ATM the only guarantee at present, such as the real-time and synchronization requirements of DDN and high service technologies. ATM will carry the IP network technology. The integration of IP and ATM is suitable for traditional telecom operators to build their public multi-service platform, and also promotes the traditional Circuit Switching Network (PSTN, ISDN, DDN, etc) the evolution to the group exchange network.
2. Centralized call processing architecture
Centralized call processing architecture, the single control module is used to process all tasks, including call establishment request, route layout information update, request interpretation and response management, management response generation, and location management. In this architecture, all calls from network interfaces are handled by a single control module in the switch. The call processing capacity of this integrated system cannot exceed 100 times per second. It is only applicable to low-end switches such as working groups. The main reason for the application of this integrated system structure is restricted by factors such as technology and economy at that time. Its main drawback is that its performance cannot be scaled down. Because when the number of network interface modules increases or the number of connections to be processed in a certain period of time increases, the control module will be overloaded, the distribution of processing capability depends on how the function segments in the control module are executed. The intelligent interface module still needs the assistance of the control module to complete the call processing task, however, with the rapid development of computer technology, network technology, and automatic control technology, the use of intelligent network interface control module can completely eliminate the bottleneck brought by the control module to call processing.
When a centralized call processing architecture is adopted and assumed that there is only one Proteus microprocessor in the control module as a computing tool, the expected performance of the system is 50 calls per second or less. The analysis shows that to process 150 calls per second, the control module must have at least two Proteus microprocessors. However, due to the limitation of hardware density, the number of processors that the control module can contain is limited. In addition, it cannot meet the needs of future development, the system architecture must be scalable, and the control module must be designed to meet the requirements of the entire load system. From the analysis, the centralized call processing structure can be obtained. With the continuous advancement of technology and the rapid development of network technology, it will be replaced by the Scalable call system structure.
3. Distributed signal call processing architecture
In the distributed signal call processing architecture, signal processing is completed in the interface module. Each interface module has its own dedicated processor capable of processing multiple tasks. In this system structure, each interface module can terminate the signal stack and participate in the call control process. The process of processing signal messages is recognized as the most time-consuming process in the call processing process. transferring this part of work to the interface module can achieve better performance and scalability, because the parallel operations of signal processing can usually improve the processing latency.
However, in the distributed signal architecture, each interface module simply converts the signal primitive from the signal stack to an internal control message, and the control module still needs to process calls, the common routing and system operations and management are also the same, which limits the performance of the architecture to a large extent. The distributed signal system can process-calls per second, and is suitable for medium-scale switches such as campus and enterprise. In the distributed signal processing architecture, the corresponding product performance will be three times the structure of centralized call processing per second, so as to achieve 150 call processing capabilities per second. The reason for adopting this architecture is as follows, first, signal tasks are easily distributed, and second, risks and high capital investment caused by distributed call control and routing architecture can be avoided.
4. Distributed call control and processing architecture
This integrated system structure is built on the distributed signal call processing architecture. It implements the call control function in a distributed manner, thus improving the call processing performance step by step. This architecture directly sends signal messages from the entry module to the ELE. Me module, eliminating the bottleneck of the control module and allowing the system to flexibly execute a large number of internal control functions. Of course, Intelligent Modules can only query routing information for call requests through the control module. Intelligent interfaces can respond to call requests destined for vswitches. The control module is used only when the destination is in another vswitch of the network. This system structure also takes into account that most call requests do not consult the routing protocol. The distributed call control processing architecture can greatly improve the scalability of the system. The cost is that distributed control requires extremely complex management of the call controller in the system, therefore, it is more difficult to implement than distributed signal processing. In addition, because the control module continues to retain the system's source information, the interface module must report decisions on local resource information to the control module at any time, which further limits the call processing performance. The distributed call control system can process up to-calls per second. It is suitable for enterprise or edge switches.
5. Distributed route call processing architecture
Assigning the path selection work to the interface module can further improve the system performance. Therefore, the control module is only used for switching the routing protocol layout. During each layout update, the control module updates the layout database data maintained by the interface module, and the routing determines the device to perform complex path computing for the given call request. Layout updates are usually carried out cyclically or only when major changes occur. These updates do not bring about a lot of work. In addition, the layout aggregation technology can also expand the routing protocol to a larger network. In this system structure, the control module no longer processes path control. The distributed routing architecture can provide 1000 calls per second, making it an ideal choice for core or edge switches in the WAN.
In order to implement the distributed routing and call architecture, researchers at the Bell laboratory used the ATLANTA chip of Lucent Technology to form an exchange hardware. In addition, they also used ERM devices that support API business control, the switch consists of an exchange matrix module, an interface module, and a control module. This technology is under development and is believed to be a major call architecture of the Next Generation Network.