Introduction to the development process of vro chips for Small and Medium-sized Enterprises

A router is a node device in an interconnected network. It is used to connect multiple networks or CIDR blocks. The router works on the layer-3 of the layer-7 protocol. Its main task is to receive packets from a network interface, based on the destination address, it is decided to forward to the next destination.

Like a computer, a vro also has a CPU. Vrouters of different levels have different CPUs. CPU is the heart of a router, both in the Middle-end router and in the high-end router. Generally, the CPU is responsible for exchanging route information, routing table searches, and packet forwarding among low-end routers. In a vro, the CPU capability directly affects the route table query time of the vro's throughput) and the route computing capability affects the network route convergence time ). In high-end routers, packet forwarding and look-up tables are usually completed by the ASIC processor, and the CPU only implements the routing protocol, computing routes, and distribution route tables. With the development of technology, many of the router's work can be implemented by hardware ASIC dedicated chip ).

The development of vro processor chips has gone through the following four stages:
1) general purpose processor
2) Embedded Processor
3) ASIC Processor
4) Network Processor

General processor stage
In the 1960s s, people used computers to act as routers. This is the prototype of the first generation of routers. One computer is integrated with multiple NICs. Multiple NICs share one processor and are interconnected through the internal BUS. the CPU is responsible for almost all route computing and data forwarding commands, at the same time, it was also responsible for the device management of the entire machine, and then gradually developed a dedicated bus, interface and operating system router.

As a general-purpose processor, because of the needs of various applications, it has a general-purpose architecture and instruction set, in order to support complex operations and easily add new features, that is: it is not specially designed for network communication. Generally, the forwarding speed is relatively slow and the scalability is poor, which is difficult to meet network requirements.

Embedded Processor stage
The biggest difference between an embedded microprocessor and a general-purpose microprocessor is that an embedded microprocessor mostly works in an application-oriented system designed by the equipment manufacturer. At present, most of the applications are specially designed to meet the requirements of high performance, low cost and low power consumption. Such as mobile communication, PDA, game machine, network communication, and other electronic products.

At present, embedded processors mainly include Am186/88, micron ex, SC-400, Power PC, Motorola 68000, MIPS, ARM Series and so on. In the 32-bit embedded processor market, there are companies such as Motorola, ARM, MIPS, TI, and Hitachi, some companies that produce general-purpose microprocessors, such as Intel, Sun, and IBM, it also produces embedded microprocessor, but is not a professional manufacturer. People are more familiar with Pentium, such as INTEL.

The first-generation router is an embedded system based on an embedded microprocessor. It has specialized circuits, interfaces, and operating systems and is a specialized equipment, it is no longer a PC based on a general-purpose microprocessor, a general-purpose interface, or a general-purpose operating system. The Cisco2501 router is a typical representative of the first-generation router. Its CPU is a MOTOROLA 68030 20 MHz processor, which is equivalent to INTEL's 80386 general-purpose processor.

Currently, a large number of embedded micro-processors are used in the low-end router products of network equipment suppliers, whether it is Cisco's general-purpose router series, or Broadband Router products used in small enterprises and homes, you can see them.

ASIC dedicated ic) processor stage
When the network speed is relatively slow, the routing and forwarding speed of the embedded processor can catch up with the data stream completely. Later, the line bandwidth is wider and the data rate is faster, the processing of embedded processors is not fast enough, so Designers Turn to ASIC. ASIC is a processing technology widely used in performance-sensitive platforms.

The invention, production, and application of routers enable rapid network development. However, in the early stages of network development, the network transmission speed is relatively low and the business volume is relatively small, network Devices in this period are generally based on CPU generic or embedded), that is, various network functions are achieved by running the relevant network operating system on the CPU. It has high flexibility. By updating the operating system, you can improve the original functions or add new functions and service features. However, the disadvantage is slow processing speed and low throughput. However, this performance was acceptable in the low-speed network environment at that time, because the speed of the router forwarding group can completely keep up with the transmission speed of the line. However, with the advancement of optical fiber and other transmission technologies, the increasing speed of network bandwidth gradually exceeds the increasing speed of CPU processing capability, this makes the "CPU + Operating System"-based routers gradually become the network bottleneck. Therefore, you need to find a way to improve the performance of network devices. In this case, network devices begin to adopt ASIC Technology. It solidified the instruction set or computing logic into the chip, and implemented all the details of the forwarding process in hardware mode, so that it can achieve a high processing speed, this can meet the performance requirements and adapt to the increasing trend of network bandwidth.

In high-end routers, packet forwarding and look-up are usually completed by the ASIC chip, and the CPU still exists. However, it only implements routing protocols, computing routes, and distribution route tables. Due to the development of technology, many work in routers can be implemented by hardware.

Network Processor stage
The advantage of ASIC is that it lacks flexibility. Once the instruction or computing logic is solidified into the chip hardware, it is difficult to modify and upgrade. To add new functions or improve performance, you have to redesign the chip. In addition, designing and manufacturing complex ASIC generally takes a long time and R & D costs are high. In addition, the application scope of the current network is constantly expanding, new services are constantly emerging, and the development of the network is not only a constant increase in bandwidth, but it is more manifested in the requirements for "Intelligent Processing", such as service quality QoS), control Security) and other services need to be classified and deep data processing to Layer 3 to layer 4th ). These service functions require both high processing speed and flexibility. Therefore, the processor needs to be able to flexibly meet different needs of various services and applications, this is hard to satisfy the ASIC Technology, which gave rise to the emergence of new processors, namely, "network processors ".
The network processor is designed to optimize packet processing. It will send data packets to the next node at wire speed. In addition, if new functions or standards are required, device manufacturers can program network processors to meet various new network applications.

It should be said that the biggest advantage of network processors over ASIC is flexibility and the development cycle is relatively short. The performance of network processors is greatly improved compared with other processors, but there is still a gap with ASIC in terms of high-speed data packet processing. In the vro field, vrouters that are at the core of a medium-sized enterprise network and at the edge of a telecom network use NP. However, there is still debate over whether the ASIC-based architecture is NP-oriented for the core routers used in telecom networks and some large domestic enterprises and institutions. The architecture of ASIC seems to have the upper hand, but everything is variable. Maybe the best way is to combine ASIC with NP and take the length of the two to create a fast and flexible core backbone router.

Summary:
In addition to general-purpose processors, vro processor chips include embedded processors, ASIC processors, and network processors. Embedded processors will appear more in low-end router products, such as home and SOHO broadband routers and VPN firewall gateways; ASIC processors and network processors will appear more in Middle-end, high-end enterprise-level, telecom-level routing routes and switches.