But if the site has an average of more than 200 requests per second, then the problem is: This is already the best Web server, what should I do? The same scenario applies to the database. To solve this problem, we need to understand the principle of "load balancing".
How Web servers do load Balancing
The most important way to do load balancing for Web servers is DNS redirection and reverse proxy, and the other way it works is similar.
We ping Baidu many times, we will find that the IP of the reply will be different, for example, the result of the first time is:
We understand the reverse proxy from the following figure:
The client sends an HTTP request message to the reverse proxy (if the site has a domain name, domain name IP is a reverse proxy server extranet IP, the reverse proxy will send the request message to a Web server randomly, the Web server will send the HTTP response message to the reverse proxy, the reverse proxy to return this message to the client. Since this is so simple, we can begin to implement a simple reverse proxy.
Install Apache and nginx servers under Linux Mint 15 and create a file index.html under the Apache 80-port document root, which reads as follows:
As long as you understand the principle of reverse proxy, more complex architecture is easy to implement.
Load Balancing of databases
For a large web site, a database system will certainly encounter the inability to afford a large number of read requests, write requests. So how do we achieve high concurrent read and write requests through load balancing?
One of the best ways to do this is to read and write a separate read-write request to a database server, send a write request to one (or more) database server, send a read request to another server (or servers), which can significantly improve response time. But one of the difficulties is that you have to keep the data in multiple database servers consistent, and you don't have to worry that many database systems have implemented this functionality. The following is an example of a schema:
In fact, there is a problem with writing a conflict, imagine the following scenario:
The system is used to store user registration information for a Web site that does not allow the user name to be the same and the user name is the only primary key, so transaction processing must be involved in a single database schema. Now in this load-balanced database schema, user A registers a user named Xiaoming with the write request assigned to DB server 1, while User B also registers the username xiaoming, and if the write request is assigned to DB Server1, no problem occurs. But what if it's assigned to DB server 2? Two DB server holds the same user information for different user names! The solution is simple, the allocation of write requests can not be a random algorithm, should use hash map, such as the registered username first letter x, write requests assigned to the DB Server2, all other write requests are assigned to DB server 1.
Another problem is that this architecture provides a great deal of flexibility for development applications, that is, this architecture does not apply to some ORM frameworks, and the solution is to add a layer--"database broker" to the architecture. For MySQL, for example, there is a solution like MySQL proxy.
Start building with 50+ products and up to 12 months usage for Elastic Compute Service
1 on 1 presale consultation
24/7 Technical Support 6 Free Tickets per Quarter Faster Response
Alibaba Cloud offers highly flexible support services tailored to meet your exact needs.
A comprehensive suite of global cloud computing services to power your business
Payment Methods We Support
