Bearer Network of the Future Internet

This work was funded by the National High Technology Research and Development Program ("863"Program) of China under Grant No. 2008AA01A301

1 Mainstream Network Technologies
For a long time, the issue of mainstream networking technology has been scarcely discussed. The communications industry has firmly believed that data packet technology will be the dominant technology of future networks, and academia has been persistently studying various new technologies and publishing research results in academic papers. With the rapid development of communication technologies—especially transmission  and switching technologies—it is no longer difficult to achieve a terabit per second transmission rate over an optical fiber, and a terabit per second switching or routing capability on a single device. And such devices can be mass-produced. Moreover, transmission resources are becoming increasingly inexpensive and powerful. On the other hand, data packet technology, which is based on statistical multiplexing theory, faces some problems. Issues such as security, and Quality of Service (QoS) greatly perplex the communications industry. In seeking solutions to these problems, the communications industry has considered timeslot switching-based network technology, and the number of voices now espousing this as dominant technology has increased. As a result, conjecture has also arisen as to whether or not data packet technology or timeslot switching-based network technology should be used in future networks.

    In the Internet environment, the network needs to support narrowband services, broadband services, fixed services, mobile services, and any combination of these. The future network should adapt to the following application scenarios: where users access narrowband services (which may be real-time voice services, non real-time data services, or a combination of these), the network should be capable of matching all narrowband services or service combinations and providing variable bandwidths to meet the demands of simultaneous real-time and non real-time services. Where users access broadband services, the network should be capable of providing a bandwidth that matches those services. Where users access a combination of above services, the network needs to provide rapid, flexible, and on-demand transmission resources within the entire network. Timeslot switching-based technology has difficulty meeting such requirements in fixed communications let alone in a mobile communications environment where network resources need to be scheduled dynamically. The problem in the above scenarios demonstrates only one side of the matter, and it can be solved by providing resources according to the maximum demand of all services; that is, 40 Mbit/s per user. But this solution leads to a great waste of social resources, and it is generally unacceptable for both operators and consumers. It is also inadvisable from the perspective of energy saving and reasonable use of resources.

    Communication between users is no longer simply point-to-point; users no longer do only one thing at a time. A common scenario might be a person is calling while surfing the Internet; or in a family, one person is surfing the Internet while another is calling, a third is watching IPTV, and yet a fourth might be downloading large-size video files using a Peer-to-Peer (P2P) system. This means current communications often involve several connections. These include point-to-point, point-to-multipoint, and multipoint-to-multipoint. In such application scenarios, it is impossible for timeslot switching-based networking technologies to dynamically schedule network resources and set up connections.

    In the above scenarios, the services are of variable rate, and this requires the network to be capable of scheduling all its resources in a flexible and efficient way. Communications between users is multi-process; that is, in any given period, communication connections are of different kinds. This requires the network to be capable of point-to-point, point-to-multipoint, and multipoint-to-multipoint communications. At present, these scenarios cannot be handled with timeslot switching-based networking technology, but can only be realized with data packet technology.

2 Transport Layer Technologies
The Telecommunication Standardization Sector of the International Telecommunication Union (ITU-T) has spent eight years (two study periods) researching Next Generation Networks (NGN) and has determined the core technology and architecture for NGN. This core technology comprises packet data technology because only it can deliver variable-rate multimedia services and meet diverse service requirements. NGN architecture consists of service layer, transport layer, and the support systems of these. One characteristic of NGN is that the service network is separate from the transport layer, and develops independently from the bearer network. Such architecture provides a broad space for service development and an open environment for delivering services to users. There are two functional modules—Network Attachment Control Functions (NACF), and Resource and Admission Control Functions (RACF)—between the service layer and transport layer. These two modules are used to connect the two layers, allowing transport layer resources to be controlled by the service layer while the two layers work independently. This lays a good technical foundation for future operational services.
As for the question of whether the transport layer comprises one or two layers, opinions differ. People who regard the transport layer as one argue that it is completely flat, and the transport network of a communication network provides end-to-end communication capability. Those who regard the transport layer as two hold the idea that the transport layer consists of two networks: packet data network and transport network. The former provides end-to-end data packet communication capability, while the latter provides point-to-point connection (dedicated line) for packet data network devices.

    It should be noted that the layers here are logic rather than physical devices. The functions of different logical layers can be integrated into one physical device. Existing transport networks mainly include optical transport networks (which are based on optical mux/demux and optical wavelength switching technologies), and Synchronous Digital Hierarchy (SDH) transport networks (which are based on SDH virtual container mux/demux, and SDH cross connection technologies). These transport networks are not capable of forwarding or switching packet data, and only provide dedicated lines for point-to-point connection between packet data network devices. Since the concept of "transport network" was coined, it has never been treated as a network but as a dedicated line of different granularities or different dimensions. Theoretically, the transport network can become a real network after the signaling system is added. But in reality, due to various dimensions of granularities, this can hardly be realized. To form a network, a unified granularity dimension has to be used. Even if the packet is used as the unified dimension of granularity, the nodes in existing transport networks cannot process data packets with current technologies. Therefore, the transport network, including SDH and Wavelength Division Multiplexing (WDM), and the packet data network will not be converged into one network in the near future.

    ITU-T’s research into NGN architecture provides useful references for the design of the future Internet.

3 Packet Data Network
There have been many packet data networks constructed on different technologies including X.25, Frame Relay (FR), Asynchronous Transfer Mode (ATM), Ethernet, and IP network. Most of these technologies were put into widespread commercial use and some are still being used in commercial networks today. In light of new demands brought about by service development and development trends in networking technologies, IP networking is now recognized as the dominant direction of future networks and will be an important part of national information infrastructure in the future. However, IP networking is not without its flaws. It must contend with two architectural bottlenecks: one is the shortage of address space (as the existing IPv4 address system cannot meet the increasing demands), the other involves network problems in security, trustworthiness, controllability, manageability, and QoS, all of which cannot be solved with existing architecture.

    The shortage of IPv4 addresses can be alleviated with dynamic address translation technology, but this solution brings with it problems with security trace, application development, and sunk costs. One feasible solution lies in IPv6. IPv6 is a widely recognized address solution for the next generation Internet. Adopting a 128-bit address coding method, the address space of IPv6 is 296 x that of IPv4. The number of address spaces is almost infinite. But the deployment of IPv6 cannot create directly unique services and market opportunities; on the contrary, it requires huge investment. Hence, industry insiders are not commercially motivated enough to deploy IPv6. The decision of one or even several participants to deploy is not really enough to initiate the market. In the past few years, all countries and participants have shown great interest, but have taken a wait-and-see attitude to IPv6. Moreover, the IPv6 international standards have almost been completed. According to its protocol and standards, IPv6 is basically an upgrade version of IPv4 except that its fundamental protocol is incompatible with IPv4. An IPv6 network, adopting the same architecture and core technologies as an IPv4 network, is also an upgrade of an IPv4 network. Indeed, IPv6 expands the address space and solves the shortage of address space forever, but its performance and other functions do not radically change when compared with IPv4. More importantly, it cannot solve security, trustworthiness, controllability, manageability, and QoS problems of the Internet.

    Because IPv6 can only solve the shortage of address space, other ways must be determined to overcome the second bottleneck. The study of next generation Internet shows two technical methods are available for Internet evolution: reformative and revolutionary. In the reformative method, the architecture of existing IP networks is not changed but improved with various technologies. With certain enhancements added to meet the bearer network’s demand, existing technologies gradually evolve into next generation networking technologies.
Using on the revolutionary method, a new network is designed based on future Internet technologies and future service application demands. This new network solves all problems that cannot be solved in existing IP networks and meets demands for future information and communications services.

    As for the controllability, manageability, operability, QoS, and security problems inherent in existing IP networks, experts favoring the reformative strategy try to amend the IP network with various technologies, and let the network meet the demands of future networks. Representatives of these technologies are Multi-Protocol Label Switching (MPLS) and its derivatives, such as Transport MPLS (T-MPLS), and Virtual Private LAN Service (VPLS), and Provider Backbone Transport (PBT) and related technologies, such as Media Access Control (MAC) in MAC. Although these technologies can partially solve the manageability, controllability, and security problems of the IP network under certain conditions, they have their own limitations. For example, the core idea of MPLS is to implement control and security similar to an ATM network, to deliver Virtual Private Network (VPN) services, and to guarantee QoS via the connections it establishes on the IP network. But being connection-oriented, MPLS suffers poor scalability as network scale and services continue to grow.

    MPLS is applicable to small networks and some VPNs. It cannot be used in a large network, especially a nationwide or worldwide network. The setup of MPLS Label Switching Path (LSP) and the distribution of labels both depend on routing protocols. On the one hand, existing and extended routing protocols simplify the implementation of MPLS; on the other, they cause problems such as convergence, loop, and network overhead. These problems cannot be neglected. In addition, to provide QoS and traffic engineering, the routing selection protocol has to provide a QoS-based routing selection function or constraint-based routing selection function, which complicates the routing selection protocol further.

    The network core also becomes much more complicated with MPLS. First, the generation, allocation, query, and mapping of labels requires the participation of the core node. Second, in order to provide MPLS services, all network devices have to support MPLS. The core node in particular has to implement complicated routing protocols and MPLS protocol in addition to high speed packet switching. The larger the network, the more complicated the processing. Third, with an increase of VPN users, the information maintained by Provider Edge (PE) routers increases accordingly, and the Border Gateway Protocol (BGP) routing table, operated and maintained by telecom operators,will become more complicated.

    MPLS, T-MPLS, and VPLS suffer serious scalability problems but do not basically solve QoS, security, and other problems. MPLS has a 12 year history and every means have been trialed to enhance its functions. It is unlikely that any breakthrough will be forthcoming. If the functions of existing IP networks were continually enhanced and their defects amended, the consequence would be that the IP networks, which were originally simple and efficient, become more complicated. Their efficiencies would decrease gradually and they would not be able to solve the root problems. Surely, the reformative approach will continue but it is unknown whether this approach can escape its current difficult situation.

    The revolutionary method completely solves the problems inherent in existing IP networks and meets demands for future information communication. It does this by innovatively redesigning a new packet data network based on future Internet technologies and service demands. To redesign such a new network, one key problem is to clarify technical requirements and work out topmost layer for the network. The new packet data network should provide solutions to problems involving work mode of the network, network control, QoS guarantee, performance management, and security.

4 New Packet Data Network
ITU-T’s Focus Group on NGN (FG NGN) has clearly specified the position of Future Packet Based Network (FPBN) in NGN architecture and has conducted research into the following: problem description, general requirements, topmost layer design, and candidate technologies. At present, the first three have been completed under the leadership of Chinese researchers and two recommendations have been given: Y.2601—General requirements of Future Packet Based Networks, and Y.2611—High level architecture of Future Packet Based Networks. These recommendations have laid a solid foundation for future research[1-2]. Now, research on specific network implementation schemes for FPBN requirements and top design (i.e. candidate technologies) has passed to Study Group 13 of ITU-T (ITU-T SG13) under the Packet Data Network (PDN). So far, China has offered a candidate scheme: Public Telecom Packet Data Network (PTDN).

    PTDN takes the non-connection-oriented transport mode as its main work mode. Theories and practice indicate that this mode offers good openness and scalability. Non-connection-oriented mode plays a critical role in the success of IP networks and is their most important feature. PTDN inherits this mode, which ensures excellent scalability.  By introducing the features of communication networks, PTDN achieves predictability, controllability, and manageability of the network. PTDN also supports connection-oriented transport mode to satisfy special scenarios.

    PTDN adopts a hierarchical network architecture, dual address mapping-based address system, ordered address structure (where addresses are assigned by region), node potential-based routing technology, and automatic multi-routing technology. All these enable PTDN to offer carrier-class protection and changeover capabilities.

    To ensure trustworthiness and security of the network, PTDN takes a series of technical measures in the data, control, and management planes. First, it divides the data plane into two areas: trustable and suspect. In the trustable area, information is transferred in a transparent way, enabling network interception, while in the suspect area, user information is transferred in a non-transparent way, guaranteeing security and integrity of user information. Second, in the control plane, Service Node Interface (SNI), User Network Interface (UNI), and Network Node Interface (NNI) are separated to secure the control plane in the network node. Third, the user in the management plane is unreachable, ensuring the node security.

    In order to implement hierarchical management and control of the network, as well as to support multiple services, PTDN adopts multiple data plane technology. PTDN supports multiple data planes, but the information in all data planes is strictly isolated and the resources of the planes are used independently. As a result, even in extreme scenarios, the resources are independent and secure. Each data plane independently performs Operation, Administration and Maintenance (OAM) to ensure its own performance.  Each data plane also has a complete signaling system.

    To ensure the reasonable use of resources, adherence to green, energy-saving principles, to guarantee QoS, and to support feasible business models, PTDN should have a complete resource management system. It should combine resource management technologies such as fair algorithm, threshold-based alarm, and overload discard to achieve precise control over network resources, and to reach the QoS class required by the service network. PTDN should also feature decentralized resource management (to ensure scalability, and to ensure its resources are configurable, predictable, and measurable), and the capability of emergency communication.
To meet the requirements of quality broadcast services, PTDN should adopt resource ensuring technology for controllable multicast. In addition, the information among multicast groups should be strictly isolated, resources of different multicast groups should be independent, delicate management should be used for multicast resources, and these multicast resources should be predictable, manageable and controllable.

5 Conclusions
With a history of over 30 years, IP networking is a kind of packet data networking that bears Internet services.

    The Internet is one of the most influential inventions of the twentieth century, and since the 1990s, it has undergone rapid development. It now interacts with globalization and has profound influence on production, daily life, scientific innovations, social services, and cultural propagation. The Internet is a driving force behind world development and change, and the transformation of human society into an information society[3]. With the worldwide popularization of the Internet, problems such as address shortage and poor security control have become increasingly serious. Therefore, the Internet’s sustainable development is greatly constrained. The Internet is now at a crossroads and is seeking a significant breakthrough to evolve into the next generation. Such a breakthrough will be born from the bearer network; that is, a new packet data network.

References
[1] ITU-T Y.2601-2006. Fundamental Characteristics and Requirements of Future Ppacket Based Networks Study Group [S]. 2006.
[2] ITU-T Y.2611-2006. High-Level Architecture of Future Packet Based Networks [S]. 2006.
[3] 蒋林涛. 电信转型和下一代网的若干问题研究 [J]. 电信工程技术与标准化, 2006, 19(1): 1-5.
JIANG Lintao. Research some questions on transformation and next generation network [J]. Telecom Engineering Technics and Standardization, 2006, 19(1): 1-5.

[Abstract] The future Internet needs to support broadband services, fixed services, mobile services and a combination of these. Such a wide range of services and multi-processes between users demands flexible and effective network-wide resource scheduling and support. Two approaches are currently being studied: reformative method and revolutionary method. The reformative method, based on existing technology, uses various techniques to achieve improvement. Revolutionary method seeks to address future business demands by completely re-designing the network, and overcoming problems that cannot be solved by current IP networks.