Load balancing is an application of cluster . Through the allocation of traffic , So as to improve the concurrent processing ability , Generally we mean web Load balancing , Now let's discuss how many ways there are for load balancing ？

First of all, we need to be clear about ： What is the purpose of load balancing ？

Distribute user requests to many servers , Solve the problem of high concurrency

1.HTTP Redirect load balancing

The load balancing process is shown in the figure above ：
1、 Request to the load balancing server ;
2、 According to the equalization algorithm , return 403, Redirect to the server actually responsible for Computing IP Go to browser ;
3、 Then the browser directly requests the server actually responsible for the calculation ;
4、 Finally, the server actually responsible for computing corresponds to the client browser ;

shortcoming
1、 The request passes through the server twice （ Load balancing server , Actual computing server ）, Added corresponding time ;
2、 Exposed the actual load of the server public network IP Go to browser , Low security ;

Java Implementation code

response.sendRedirct("http://127.0.0.1/service?param=a");

2.DNS Load balancing

The load balancing process is shown in the figure above
1、 User requests domain name , Ask to DNS The server ;
2、DNS The server returns parsed IP Address to client , It is not exposed to the browser for redirection ;
3、 The client gets the returned server in charge of calculation IP, Request server ;
4、 The computing server returns the corresponding information ;

improvement
1、 You don't need to request the load every time IP, It can be cached , Reuse , Provide performance ;
2、DNS There is no need to expose the actual computing server IP（ It is not exposed in the browser by redirection , Instead, I did secondary load balancing , Intranet IP It won't come out ）, The security is slightly better ;

This method is adopted for multi activity in different places , One IP Resolve to different regions IP, Realize the load balancing of the first layer , Then based on the region IP Do secondary load balancing ;

3. Reverse agent load balancing

The load balancing process is shown in the figure above
1、 The client initiates a request to the load balancing server , The load balancing server gets the load according to the algorithm IP;
2、 The load balancing server constructs the request , Computing server requesting intranet load ;
3、 The computing server returns the corresponding results to the load balancing server ;
4、 The load balancing server returns the corresponding results to the client ;
such as nginx、apache

shortcoming
1、 be based on HTTP Load balancing done by layer , It is a relatively heavy agreement , It's a little inefficient ;
2、 It is generally suitable for smaller clusters ,10 + scale ;

4.IP Layer load balancing

On the network layer IP Address to replace , Don't need to HTTP Work on the layer , Directly in the operating system kernel IP Replace address in packet , Efficient than HTTP Layer reverse proxy high ;

The load balancing process is shown in the figure above
1、 The client requests the load balancing server ;
2、 Load balancing server modification purpose IP For intranet machines IP;
3、 The calculation of the intranet machine is completed , Corresponding IP Change to server balanced server IP Internal network address ;
4、 The load balancing server modifies the corresponding IP For your own Internet IP, Return the result to the client ;

shortcoming
Requests and responses need to go through the load balancing server IP Layer replacement , The corresponding data will become the bottleneck in the later stage ;

5. Data link layer load balancing

Solve the problem of excessive data and low efficiency , By modifying the data link layer mac Address ,IP Using virtual IP, To achieve load balancing ;

The load balancing process is shown in the figure above
1、 The client requests the load balancing server ;
2、 Load balancing server replacement mac The address is the computing server ,IP Load balancing server IP;
3、 The computing server directly sends the corresponding data to the client ;

This load balancing method has the highest throughput , Large Internet companies adopt this load balancing method ;
LVS Load balancing (Linux Virtual Server) It's a combination of IP Load balancing method of layer and data link ,IPVS Is its specific implementation module ;Linux These two load balancing methods can be realized through configuration .