How To Build An Excellent Large-scale Network Architecture?

- Aug 12, 2020-

1:Small network architecture

Let's first think about what the network generally talks about. We start from an early age, so let's talk about a small typical network structure first.


For example, a small company may have a hundred people, or we may have a dormitory, which can actually be said to be a small network. Of course, our dormitory uses only two or three people on the network, and more dorms may have seven or eight people.


Small companies have more than one hundred personal networks. Its structure is relatively simple. Generally speaking, add a router and this router can also serve as (01:40) CP, and then build a file server, connect to printers, and connect to terminal computers. Add another wireless (using wireless for management), and then add a firewall outside, and you can also buy a device behind the firewall. The firewall is also saved. This is a small network.

8644ebf81a4c510fdd7e93cf37708b28d52aa55f

The medium-sized network is probably designed in this way, but it must be a little more complicated than what we are talking about now. Roughly in a typical small network, adding things to the top, more layers, one layer becomes three layers.

2:Large-scale network architecture

Next, we will continue to talk about large-scale networks. Access to more than one thousand devices can be regarded as large-scale networks. Large-scale network access is more complicated than medium-sized networks. If routers and firewalls are connected at the same time, the switch may not support NAT (Network Address Translation).


The external network access of the large-scale network structure is provided by the operator. Generally, two access operators are selected, because if the line is down, the network may also be down, so at least two should be connected.


Then the outer layer of the connected device is usually connected to a firewall, some are connected to a router behind the firewall, and some firewalls and routers are together, and no separate router is connected, and the firewall is used directly.

5bafa40f4bfbfbed824ac6072fd95933afc31f61.jpg

What is the use of core switches?


It has many functions, you can plug in many boards. If you plug in the firewall board, it has the function of a firewall. If you plug in the router, it has the function of the router. The function of NAT is also possible, just say this You can buy the board or not.

The backbone network is mainly core and data.

Large-scale networks have dedicated data centers, and data centers are more complicated to build. They are mainly groups of servers, and servers are generally placed in data centers.


Generally speaking, a large-scale network has a core switch, and then it is connected to our backbone core switch. The switch in the database is still a little different from our usual switch, because the switch usually sends packages called 100G or dozens of G, in fact It is different when transferring large data files.


Then the backbone network is connected to the wide area network. You see two WANs, one is the operator and the other is the WAN. The WAN is mainly used as a dedicated line, because large networks have branches. If the branches are connected to other places, they need to use this dedicated line, so they will be connected exclusively through the WAN. There is still a difference between dedicated lines and Internet access. Internet access is mainly to access the Internet.


The dedicated line of the WAN serves to connect the parks in different places, and the connection is the control of the access between the buildings. For example, there are many floors and many buildings in a factory or park. How does it connect when it is connected?


In addition, because it is all wireless control now, wireless also has special wireless access. This wireless will be controlled by the controller, which is usually hung under the backbone network core.


But if the network is so complicated, when we want to maintain it, we need a special software to control it as a network management platform. The network management platform monitors its traffic or abnormal conditions.


In fact, the structure of a large-scale network is actually not complicated. The first is a small-scale network of a small company. At the beginning, there is only one switch and only 100 people. The small-scale network is equivalent to accessing one floor. Then the company and personnel develop and grow into a medium-sized network. A medium-sized backbone network is connected. When the company becomes more developed, it has to grow to more than a thousand devices, it will need to add other things, add a branch, and then connect to the remote network, because the company is relatively large, you need to add a wireless controller to the backbone network, and then also Need to build a data center, because the network is relatively large, there must be a network management platform, so another network management platform is added.


In fact, its design is not that complicated, it is equivalent to a step-by-step development. If you remove all the side branches, you will first look at a backbone network, which is a three-tier structure.


The so-called two-layer structure means that the connected switch is the core switch, and the advantage is that it is very fast. The following is the server storage data center, you can use ARP technology, generally this core is a single row, and then connected by links in the middle.

80cb39dbb6fd5266a1326ef4fd31392ed50736e7.jpg
Then look at the three-tier structure. What is the difference between the third layer and the second layer, the only advantage is that it can reduce the investment of the core switch. Because the core switch is more expensive, the disadvantage is that the conversion delay will be slightly larger. Of course, this delay can be ignored. Now the network performance is very strong, and the conversion time is also very short, all in milliseconds. You can't even feel it Delay.

060828381f30e92428ca2eba1a21c0031c95f7fa.jpg

Therefore, the advantage of the three-layer structure is that it can save costs, because the cost of the core switch is much higher than that of the aggregation switch. A sink switch is about tens of thousands of dollars, and a core switch requires hundreds of thousands.


So whether to choose the third or second floor, you can calculate according to your scale, or you can choose the second floor if you don't need money. If investment is considered, three layers are used.