|
At the very beginning, let's talk about what load balancing is. Load balancing is to distribute a batch of requests to different backends for corresponding processing according to the content of the requests, thereby providing load sharing, master-slave switching and other functions. Different load balancing software has different traffic distribution algorithms. Today, we will compare the two most mainstream load balancing software on the market to see their respective advantages and disadvantages, as well as how to cooperate in many cases. 【Fourth and Seventh Floors】 First, let’s talk about the difference between four-layer and seven-layer; Layer 4 load balancing refers to IP+port load balancing;
Layer 7 load balancing refers to load balancing based on application layer information such as WEB requests and URLs.
Of course, by the same token, there is also Layer 2 load balancing based on MAC addresses and Layer 3 load balancing based on IP addresses.
Four-layer load balancing mainly analyzes the IP layer and TCP/UDP layer.
Seven-layer load balancing requires analysis of application layer protocols, such as HTTP protocol, URL, cookies and other information. 【About LVS】
LVS has a strong load capacity because its working mode is very simple. It only distributes requests. Moreover, it works at the fourth layer without traffic and has the highest efficiency.
LVS is at layer 4 and can perform load balancing for almost all applications.
However, LVS is not sensitive to faulty backends. For example, in DR mode, if a backend server is not configured with VIP, part of the requested data will be directly lost.
In addition, LVS has high requirements for the stability of the network environment. If the request fails, it can only rely on the retry mechanism of the front-end application itself, and the load balancing will not resend the request.
Moreover, LVS is also limited by the network architecture. At the beginning of the design, it is necessary to consider whether the network architecture meets the prerequisites of the LVS load. 【About nginx】
Similarly, nginx can also be used for load balancing, but because nginx needs to establish connections to both the source and destination ends, the speed of processing traffic is limited by a series of configurations such as machine I/O, CPU memory, etc., so nginx's load capacity is relatively poor.
The installation and configuration of nginx are relatively simple. Compared with LVS, nginx does not require a very strict network architecture, as long as the network can be connected.
And nginx's own retry mechanism can ensure that if a request fails to be sent, it will be re-sent to a healthy backend.
However, since nginx does not have a ready-made hot standby mechanism, there is a single point of failure problem and it usually needs to be used with keepalived.
However, as an application layer load balancer (later, after the introduction of the stream module, the fourth layer is also supported), nginx can provide load sharing, reserve switching, HTTPS writing, bandwidth speed limit, hiding the real IP, hiding the real port, shielding attacks and other capabilities, which LVS cannot provide. 【contrast】
LVS and nginx are both very mainstream load balancing methods. They each have their own advantages and disadvantages. In the production environment, you need to choose according to their characteristics. | LVS | Nginx | | Four layers | Fourth floor/seventh floor | | Load resistance | powerful | weak | | Configurability | Low configurability It also reduces the possibility of human error | High configurability Some advanced features can be configured | | stability | High stability There is a complete dual-machine hot standby solution | Low stability, single machine failure No ready-made dual-machine hot standby solution | | Network architecture dependencies | Strong Dependence Very dependent on network architecture design Of course, a simpler NAT method can be used to solve this problem | No dependencies | | Network stability dependency | rely When a data packet is distributed to a bad backend, it will not be redistributed and will directly return an error. | No dependency After a packet is distributed to a bad backend and returns an error, it will try to redistribute it to a healthy backend | | Network traffic | Only request traffic passes through the lvs network, and response traffic is returned by the backend server's network. FULL_NAT is the same as Nginx. | All request and response traffic will go through nginx | | Host performance requirements | Lower requirements LVS only distributes requests, and traffic does not go out of it, so the bottleneck is only limited by network bandwidth and network card performance. | High requirements Because nginx needs to establish separate connections to both the source and destination, and some data packet parsing is involved in the middle, it depends on the I/O performance and CPU memory of the host machine. | | Forwarding method | Synchronous forwarding After receiving the request, the LVS server immediately redirects to a backend server, and the client directly establishes a connection with the backend server. | Asynchronous forwarding While maintaining the client connection, a new request with the same content is sent to the backend. After the backend returns the result, nginx returns it to the client. | | other | | Support rewrite rules: can divide http requests into different backend server groups according to different domain names and URLs. Save bandwidth: Support gzip compression and add a header for local browser cache. |
【Combination of the two】
In terms of usage, the strategy generally adopted at the front end should be LVS, that is, the DNS should point to the LVS balancer. The main reason is that although nginx is powerful, when the scale of the back-end server is large, the network bandwidth of nginx becomes a huge bottleneck.
However, when LVS is used as a load balancer, once the backend server that receives the request has a problem, the request will fail.
Therefore, in many cases, nginx will be used as a node for LVS to perform load balancing. This can not only avoid serious bandwidth bottlenecks caused by nginx's performance, but also use nginx's error retransmission to avoid LVS one-shot transactions, and also use various advanced features of nginx, including https offloading and message header modification. The above is the detailed content of the advantages and disadvantages of nginx and lvs and the suitable usage environment. For more information about the comparison between nginx and lvs, please pay attention to other related articles on 123WORDPRESS.COM! You may also be interested in:- A brief discussion on two current limiting methods in Nginx
- Detailed explanation of how to use Nginx + consul + upsync to achieve dynamic load balancing
- Detailed tutorial on configuring nginx for https encrypted access
- How to view nginx configuration file path and resource file path
- Analysis of the process of implementing Nginx+Tomcat cluster under Windwos
- Configuration operations of different projects in the secondary directory after nginx configures the domain name
|
