About Server Load balancer

DNS round robin
DNS round robin refers to interpreting the same domain name to different IP addresses and randomly using a host technology. However, it has obvious disadvantages: Once a server fails, even if the DNS settings are modified in time, it still takes enough time (refresh time) to take effect. During this period, the client computer that saves the address of the faulty server cannot access the server normally. DNS load balancing uses a simple round-robin LoadAlgorithm, Cannot distinguish between servers, cannot reflect the current running status of the server, cannot allocate more requests to servers with better performance, or even the customer requests are concentrated on a server.

F5 BIG-IP
Introduction: F5 Networks is a famous Hardware load balancing switch. Supports layer-4 and layer-7 hardware switching. Different models of performance, BIG-IP 6400 can support 8 million concurrent connections, a lower model can support more than 4 million concurrent connections. High performance, but expensive.
Price: BIG-IP 6400 price at around 0.16 million yuan.
URL: http://www.f5.com.cn/(China) http://www.f5.com/(global)

LVS (Linux virtual server)
Introduction: layer-4 software exchange. LVS performs layer-4 Switching in the Linux kernel. It only takes 128 bytes to record a connection information and does not involve file handle operations. Therefore, there is no limit of 65535 file handles. LVS features high performance and can support 100 ~ 4 million concurrent connections.
Price: free, open-source
URL: http://zh.linuxvirtualserver.org/

l7sw (layer7 switching)
Description: layer-7 Software switching. This is a new load balancing software similar to LVS. I have never applied it in real time and the performance is unknown, so I will not evaluate it. This is an introduction to layer7 switching is driving a low-level Engine Using Networking Design to speed-up forwarding of data stream. implementation in this project is split into a userspace daemon and a low-level kernelspace forwarding engine. userspace daemon is responsible for scheduling and switching decisions. kernelspace forwarding engine is responsible for forwarding stream and using TCP-splicing scheme. TCP-splicing is the postponement of the connection between the client and the server in order to obtain sufficient information to make a routing demo. this project is close to Linux virtual server project since lot of discusions on this topics have been made online and offline LVS project.
price: free, open source
URL: http://www.linux-l7sw.org/

Haproxy
Introduction: layer-7 Software exchange and reverse proxy server. Currently, it does not support virtual hosts, but its configuration is simple and has a very good server health check function. When the proxy's backend server fails, haproxy will automatically remove the server, after the fault is restored, the server is automatically added. In addition, haproxy supports hot standby. I have used it for a while and can support 2 ~ 30 thousand concurrent connections. Now I use it for normal small-concurrency load balancing, mainly using its server health check function.
Price: free, open-source
URL: http://haproxy.1wt.eu/

Nginx
Introduction: layer-7 Software exchange and reverse proxy server. It supports Virtual Hosts with high configuration and load balancing by URL. I have been using it for a long time. It can support about 3 ~ 50 thousand concurrent connections.
Price: free, open-source
URL: http://wiki.codemongers.com/NginxChs (Chinese wiki)