Architecture: load balancing

introduction

• This paper introduces several load balancing architectures and algorithms .

Overall strategy

• DNS Load balancing is used to achieve geographic level load balancing ;

• Hardware load balancing is used to achieve cluster level load balancing ;

• Software load balancing is used to achieve machine level load balancing ;

Hardware load balancing

• It's like a router 、 Switch ;

• advantage ：

  • Support various load balancing algorithms ;

  • Support 100 Million concurrent （ Generally, software load balancing also supports 10 Million concurrent ）;

  • Many devices support load balancing at the same time 、 A firewall 、 prevent DDOS Attack function ;

• shortcoming ：

  • The high cost ;

Software load balancing

• advantage ：

  • cheap ;

  • Simple maintenance and deployment （ install Ngnix Software can ）;

• shortcoming ：

  • Concurrency is far less than hardware load balancing ,Linux Install a... On the server Nginx Maybe we can get to 5 Ten thousand seconds ;

  • Generally, there is no firewall 、 prevent DDOS Attack and other functions ;

DNS Load balancing

• Achieve geographic level load balancing ;

• DNS The server resolves the domain name to the one closest to the user's host IP Address , Improve access speed ;

• shortcoming ：

  • DNS Cache cannot be updated in time , It is possible to locate a removed host ;

  • Except for mapping IP Address , No other load balancing algorithms and strategies are provided ;

Equalization algorithm

• polling ：

  • After the load balancing system receives the request , It is allocated to the server in turn ;

  • Method is simple , The state of the machine is not considered ;

• Weighted polling ：

  • among 32 The probability of a nuclear machine is assigned to 16 Twice the probability of a nuclear machine ;

  • Considering the machine performance , But it cannot be dynamically adjusted according to the machine state ;

• Load minimum priority ：

  • Determine the load balancing strategy according to its specific state ;

  • consider ： Number of machine connections 、 Mechanical HTTP The number of connections 、CPU Occupancy rate 、IO Occupancy rate ;

• The best performance is preferred ：

  • Determine the load balancing strategy according to its specific state ;

  • consider ： Server response time ;

• Hash：

  • To the source IP Address hash Decide which server the task is assigned to ;

  • Yes session ID Conduct hash Decide which server the task is assigned to , It can ensure that the packets of the same session are sent to the same server for processing ;

source: //changsiyuan.github.io/2018/07/15/2018-7-15-Architecture-4

Share & Looking at