New Network "Nuclear Energy": core routers in the new environment (1)

With the rise of IP services, more traditional services are being integrated into the IP backbone network. At the same time, users also want to have the reliability and validity of services similar to PSTN. In order to become a new public network, an IP network must combine the universal connectivity of the Internet and the reliability and security of the private network. Therefore, the core router must face new challenges in the new environment.

From the perspective of the rapid growth of user requirements and new applications, the increase of mainstream applications, and the growth of important services, IP technology is rapidly becoming the choice of public basic data networks. The IP networks owned by telecom operators must combine the universal connectivity of the Internet and the reliability and security of private networks to become a new type of public network. In this situation, operators have increasingly strict requirements on core routers. At the same time, the expectations for reducing capital costs and operating costs are becoming increasingly urgent.

New core of the new network

What aspects of the core router will carriers focus on? In an interview with reporters, an expert of China Telecom first stressed that the core router is in the network center, and it must pay more attention to performance than to functions. Including high throughput, forwarding performance, fast convergence of routes, and stable Restart (Graceful Restart) are all important links of operators. A distributed architecture with processing capabilities on each line card is essential for the current core router. On the other hand, it is the degree of support for MPLS. Including MPLS Traffic Engineering, MPLS load balancing mechanisms, and so on. Another point is IPv6 support, which requires dual-stack support and hardware forwarding of IPv6 packets.

IP networks are carrying more and more services, including new value-added services. This is an irreversible trend. In addition to traditional Internet services, services such as frame relay, ATM, voice, video, and VPN are all carried out on IP networks, and IP networks are becoming telecom-level basic networks. Therefore, the network puts forward new requirements for core routers in QoS, MPLS, scalability, and IPv6.

QoS when an IP network becomes a carrier-level basic network, the psychological expectation of ordinary users is very high. Therefore, a stable, reliable and secure IP network is crucial for operators, which requires QoS Assurance. Now, the ATM over IP Address has become the actual situation. For example, China Unicom's seven regional centers are interconnected through IP/MPLS, and its core QoS is provided by IP/MPLS networks. So in the long run, more and more QoS will be provided by the IP network, and the core router must comply with this trend.

MPLS is the basis for cost-effective, highly reliable multi-service IP networks. Through MPLS, service providers and enterprises can improve bandwidth efficiency and scalability, reduce operating and management expenses, and provide reliable services. MPLS is also an important technology to support new IP services such as layer-4 VPN. It can also support existing leased lines, frame relay, and ATM services through layer-2 VPN and Pseudo do Wires. The core router must also be able to well support the deployment requirements of MPLS.

Scalable telecommunications networks must be able to expand with the increase in the number of users. This scalability will be very huge to the telecom level. Some technologies and products that can effectively achieve scalability in a local or small scope cannot meet the requirements at a large scale. This is the challenge of changing the quantity to qualitative change. On the other hand, scalability increases with the increasing demand of applications, which requires smooth expansion and cannot be skipped. Some experts believe that router equipment is a bottleneck in today's development of IP networks. Currently, the optical transmission technology is very advanced, and the bandwidth of common systems is several hundred GB, while the router port is only 10 Gb at present. The speed supported by the router technology lags far behind the transmission speed. In this case, most of the development of vrouters is in a leap. vro manufacturers have launched a new large-capacity product within two to three years, simply eliminating old devices. This kind of non-smooth expansion is detrimental to the investment protection of operators. Therefore, the smooth scalability of core routers is not only a requirement for expanding the network size, but also a need for investment protection by operators.

Due to IPv6 addresses, 3G and NGN will be developed based on IPv6 in the future. To ipvpp5, all the signals must be applied to IPv6. To ipvpp6, even the voice traffic must be carried on IPv6. As for NGN, if it does not have enough IP addresses, it cannot be carried out, and it must be applied on IPv6. The core router is crucial to IPv6 support.

New core features

What technical means should we rely on to meet network requirements? You can briefly describe the core hardware, core software, high availability, scalability, security, and IPv6 support.

Core hardware

The core hardware is a key part of the router. Vro's core hardware is based on three forms: CPU, ASIC, and NP. CPU-based core routers have been eliminated, while NP-based routers have not yet occupied the mainstream market. At present, the mainstream core routers are based on ASIC.

In general, these three forms are CPU chips. However, ASIC is a proprietary CPU and is specially optimized for an application. The ASIC applied on the core router is programmable to meet complex network requirements. NP is not as fast and efficient as ASIC in processing core network applications, and has many EDGE network applications. In short, the core router must be hardware-based, and mainstream products are ASIC-based.

Core Software

Traditional routers use a simple microkernel, Which is superimposed step by step to add new features. Because UNIX systems are more scalable than any other proprietary system, the core software of the core router is basically based on the UNIX operating system. Core software not only provides powerful operating systems, but also supports a wide range of IP service toolkit. Ensure efficient and predictable basic IP services. With this scalable and highly available network, you can flexibly segment traffic, create a unique application environment, or deploy income-generating IP services.

In short, the core software should have the following features: modularization to provide good fault recovery capabilities; Ensure that IPv6 and other new functions can be easily integrated; security, the software can run in the protected memory, ensure that functions do not interfere with each other; support a wide range of services, and provide a guaranteed experience for various types of end users.

High Availability

The Inherent trust and understanding of the reliability of the Group network is crucial to the development and growth of IP networking. This is not an easy task because the initial design requirements for Internet protocol and IP network components do not list carrier-level availability as a reasonable target. Therefore, the high availability of the core router is particularly important.

To achieve high availability, from the hardware point of view, we need to have a good system architecture and a variety of redundancy. Key components such as the routing engine and Switching Matrix must be redundant. In terms of software, it must be strong. In addition, when it encounters network adjustments such as hardware replacement, system upgrade, board Addition, and link change, the software must be able to ensure that the entire network business is not affected by local adjustment, so that the entire network shows high availability. It must ensure that the routing engine is switched without packet loss. If the primary engine fails, no packet loss occurs when switching to the secondary engine. Otherwise, hardware redundancy is meaningless and false redundancy occurs.

In addition, stable restart must be ensured. Generally, when a route is restarted, the resulting route re-calculation and route update within the network will consume processing resources, and unexpected network behavior in the form of a black hole or instantaneous forwarding loop may occur. Stable restart will avoid this situation. However, stable restart must be implemented at the IGP, BGP, LSP, LDP, RSVP, and other protocols to avoid service impact, so as to meet the needs of telecom-level networks, it also supports MPLS. Finally, high availability also requires fast re-routing (FRR) and BFD (Bi-direction Detection) to support and ensure. FRR can be implemented on MPLS to achieve 50 ms switching for fast protection. The speed can be matched with the optical transmission system.

Scalability

As mentioned above, with the development of optical transmission channels, when the router capacity is insufficient, today's approach is to replace the original with a larger router. After a while, we will continue to upgrade and add code. Generally, two or three years is a replacement period, which means investment protection is not good for users. A better way is to add a new router when there is a shortage of router capacity, so that the router grows up slowly. This is the routing matrix technology. It ensures economic and smooth scalability, And the routers deployed in the past can continue to be used without making major adjustments to the network.

Today's core routers must have the routing matrix capability to meet network scalability requirements and fully protect users' investment. Therefore, product release should focus on long-term network expansion requirements and be ahead of schedule.

Security

The security of the network should be guaranteed in the router architecture. Generally, the architecture of a router has two planes: the routing plane and the forwarding plane. If the attack is targeted at the routing plane, a hardware barrier should be built between the attack and the routing plane for filtering and processing to minimize the attack threat. Hardware barrier protection has become a widely accepted technical point of view. Using software barrier to protect the routing plane may become a critical point of attack breakthrough.

IPv6 support

IPv6 forwarding does not depend on software, but must rely on hardware. Because IPv6 and IPv4 are in different times, QoS and security for IPv6 must be in place in one step, and the starting point must be high. A wide range of complex network services will not allow IPv6 to develop from the ground up like IPv4. If the product is not prepared for IPv6 from the beginning, but upgraded by patching, it may encounter application difficulties in the future and cannot meet the long-term requirements for IPv6.