Composition and daily work of modular Gigabit Switches

With the development of enterprise network applications, people are no longer satisfied with the desktop network speed of 100 MB. Many people are looking for faster LAN connection technologies, using Gigabit Ethernet or faster 10-Gigabit technology. Such a change not only occurs in the backbone of the campus network, but also in each wiring room. The IT management personnel of the enterprise hope to provide 10 times the previous network connection for each desktop user.

There are many reasons to choose, and understanding is the basis of choice. Do you know about GE Ethernet switches? In this topic, we chose this proposition, hoping to take you inside the Gigabit Switch to see how it works, how it is made up, and why it is so different. As these problems are solved, the foundation of your products will be stronger.

When designing switches, device manufacturers usually need different settings to meet customers' requirements. These requirements generally include the following:

1. Stacking is supported for full scalability;

2. Elastic backbone scalability can be achieved through the frame structure;

3. It can be integrated with server devices to achieve integration of computing and network;

4. Small and Medium Enterprises require a fixed port or a switch that can work independently;

5. Non-managed devices will become more popular for small companies without IT departments.

There are also other requirements, such as reducing design, production, and assembly costs. Many of these requirements are conflicting, but they actually exist. The method to satisfy them is achieved through different types of switch products, and the progressive semiconductor technology makes the design and manufacture of switches enter the highly modular and integrated stage, it provides technical convenience for designing different types of switches. Next, we will show the design principles and architecture of different types of switches one by one, to help you determine what type of switch is what you really need.

Stackable Switch　

These switches are often used to connect users to high-speed campus network backbone. Generally, they have some so-called high-speed "user" ports and are highly scalable. When the data is aggregated from these ports, it will be transmitted from the higher-speed data uplink path to achieve data interaction with the central server, ip pbx and other devices. A stackable switch is usually placed in an enterprise's wiring room or data room. It can adapt to the increasing network. If a new user is added to the network, the Administrator simply places a new device on the original device, connect all switches through an external "stack" interface. In fact, just like you have developed a new and larger switch, you can easily manage it with the original switch, but the capacity increases.

A high-speed stacked port is a basic feature of a stackable switch.

Frame Type Switch　

This type of switch applies to high-performance backbone network layers. The system has many slots and can be used to connect many redundant network interface cards, including the system bus, power supply, and security modules. The entire system can be configured as a high-redundancy system that can still provide available network services in the event of uncontrollable disasters or faults. Performance is an important indicator for frame devices, but more importantly, it is the high reliability and availability of the system. The failure rate is usually required to be within an average annual downtime of no more than 5 minutes.

Figure 1 stackable Gigabit Switch structure of a 24g Port

If a stackable switch in Figure 1 is connected to an exchange backboard with a larger switching capacity switching array chip, coupled with redundant power supply and management module design, it constitutes a core Frame Switch. The high-speed switching Array Chip structure 2 is shown.

Figure 2

Switch Integrating Data Server

This type of switch can also be called the application server of the integrated switch. The blade servers we usually mention belong to this type of product. They are connected through the high-speed data channel to the frame of the blade server, and the frame of the blade server is like such a switch. The switch backplane of the blade server is usually a gigabit level. It allows servers to interconnect at extremely high speed, which makes high-speed and distributed computing possible. These are the development directions of network computing.

Like a frame switch, the backplane of the blade server is also a high-speed backplane, which can be seen as a switch. Figure 3 is an exchange chip that can be used for the blade server backboard. It can also be used for general independent or stackable switch design.

Figure 3

Independent fixed port) Switch　

This is the most common switch, which provides low-cost network solutions for Growth Enterprises. The system has fixed interfaces that cannot be expanded). When enterprise growth needs to expand the number of switch ports, the network administrator must purchase some more switches.

In general, the composition of a vswitch is similar to that of a common computer. The switch has its own CPU, memory, and external memory FLASH functions similar to the computer's hard disk), and the boot memory is similar to the computer's BIOS) and an important switch chip and network interface chip are similar to NICs in a computer ). 4.

Figure 4Independent Switch construction

Unlike ordinary computers, most network interface chips perform storage and forwarding operations through switching chips instead of the system CPU after receiving data. The independent fixed port mentioned above) vswitches and non-managed vswitches are in this structure. The difference is that an independent fixed port) A vswitch only slightly outweighs a more advanced vswitch in terms of port scalability, but according to the design, in the system, you can use a very advanced CPU and a large amount of memory. In this way, multi-layer switches can use the system's CPU to process policies such as routing and network security to implement advanced network applications. Many switches use general-purpose CPUs. Of course, they are not common Intel, but some of them are also some of the best of them.

In Figure 4, BCM5464 is a 4-port Gigabit electrical Transceiver With Integrated copper/optical media interfaces. This is what we call a network interface chip. It needs to work with the physical layer defined by optical fiber or copper wire) to form an interface for sending and receiving Ethernet raw signals. Then, it connects to a 12-port Gigabit switching chip. The 12-channel Gigabit signal transmitted by the three BCM5464 is switched in the BCM5691 chip, the corresponding MAC Learning Mechanism and ing table are created, and then forwarded to the destination port through the corresponding BCM5464.

For non-managed switches, because the demand is simpler, you can use less memory and lower-consumption and lower-cost CPUs during design. Commonly, there are various CPUs in the ARM kernel. This simple design only serves to start the system and Download Simple Control commands to the switching chip, so it is impossible to complete high-level exchange.