Combined equalizing method based on large-scale website

Abstract

The invention discloses a combined equalizing method based on a large-scale website. The combined equalizing method comprises the following steps: classifying information of the comprehensive large-scale website into data files and database data; equalizing the data files through LVS.DR (Linux Virtual Server Direct Routing); performing IP binding on an IPVS (IP Virtual Server) module to achieve an IP-level equilibrium strategy; meanwhile locally accessing files in an independent folder through an NGINX, so as to return the data files immediately, directly return the data to a client, and solve the network bottleneck problem of an equalizing device; performing reverse proxy on the database data by NGINX and equalizing the database data to a corresponding application server, and providing data support through a distributed database by the application server. The session value is not required to be shared by shared memory, and the automatic sticky connection to the corresponding application server is only required according to the session, so as to wholly reduce the possible bottleneck caused by session sharing.

Description

technical field [0001] The invention relates to a combination balance technology of network applications, in particular to a method based on large-scale website combination balance. Background technique [0002] The mainstream balancing technologies in the prior art mainly include the following five types: (1) HTTP redirection load balancing, which uses 301 Moved Permanently (permanent relocation) in the HTTP protocol to allow the client to send a new load through the returned Location again. The request; by returning different Locations to access different application servers, so as to play a balancing role. ⑵DNS domain name resolution load balancing, which refers to the use of DNS (domain name) server, here, usually refers to the secondary DNS server, such as www.myweb.com, you can set the domain name server to myweb.com through the primary domain name, and then myweb.com provides artificial domain name service, and adds multiple records in www.myweb.com, so that the DNS ...

AI Technical Summary

Problems solved by technology

[0003] The problems existing in the above existing balancing technologies are: (1) HTTP redirection load balancing is a cross-Internet balancing solution, which is rarely used as a balancing solution in actual large-scale websites, because it cannot evaluate the validity of the transferred connections , and do not know the user situation of the application server, and need more Internet IP resources, it is difficult to effectively manage

⑵DNS domain name resolution load balancing is also a cross-Internet balancing solution, which is rarely used in actual large-scale websites. This is basically a method for group companies to integrate various sub-server websites around the world. Compared with these websites, it can Relatively independent, and this also requires more Internet external network IP resources, it is difficult to effectively manage

(3) Reverse proxy load balancing can form a sticky balance for SESSION through the upper layer protocol, so that the balance is more suitable for interactivity, but because it must pass through proxy load balancing, it is easy to make it itself a bottleneck

⑷IP load balancing is more efficient than reverse proxy load balancing, but it is not suitable to form sticky balance through SESSION, only through IP address. Caching mechanism can form a bottleneck

⑸ Data link layer load balancing is the most efficient, because it works at the data link layer, and the data can reach the customer directly when it comes back, which is very suitable for static data, but it also cannot obtain the information of SESSION, so that it In the actual balance, the shared cache mechanism is also used to complete the balance, but the shared cache is also easy to form a bottleneck, especially in the case of very ordinary interactions

Images