Byte flybook Human Resource Kit three sides

Byte flybook human resource suite has three sides

hashmap Why red and black trees ？

Red black tree structure ： First of all, red and black trees are also AVL Balanced binary trees , The nodes are composed of black and red nodes . None of the red black trees is twice as far from the root node to the leaf node as the shortest root node to the leaf node , And the number of black nodes on each path from root node to leaf node is the same , So it is a weakly balanced binary tree , be relative to AVL For trees , Red and black trees have fewer rotations , For search 、 Insert 、 Delete multiple operations with red black tree .

Nginx Load balancing strategy ：

In a server cluster ,Nginx Acting as a proxy server （ That is, reverse proxy ）, To avoid too much pressure on a single server , Forward requests from users to different servers .

polling （ Default ）

The most basic configuration method , It is upstream Module default load balancing , Each request is allocated to a different back-end server in chronological order

fail_timeout： And max_fails Use a combination of

max_fails： Set in the fail_timeout The maximum number of failures in the time set by the parameter , If at this time , All requests to the server failed , So the server is supposed to be down

fail_time： The length of time a server is considered down , The default is 10s

backup： Mark this server as a standby server , When the main service stops , The request will be sent to this server

down： The tag server does not participate in load balancing

The weight （weight）

Specify the probability of polling based on the polling strategy ,weight The default value is 1,weight The value of is proportional to the access ratio

The higher the weight, the more requests to process . This strategy can be used with least_conn and ip_hash Use a combination of . This strategy is more suitable for the situation that the hardware configuration of the server is quite different

ip_hash

Ensure that each visitor has regular access to a back-end server

stay nginx edition 1.3.1 Before , Can't be in ip_hash Use weights （weight）. ip_hash Cannot be associated with backup Use at the same time . This strategy is suitable for stateful Services , such as session. When a server needs to be removed , Must be manual down fall .

least_conn

Forward requests to back-end servers with fewer connections . The polling algorithm is to forward the requests to each back end on average , Make them roughly the same load ; however , Some requests take a long time , It will lead to high load on the back end . In this case ,least_conn This way can achieve better load balancing effect

This load balancing strategy is suitable for server overload caused by different request processing time

Third party strategy

The third-party load balancing policy requires the installation of third-party plug-ins （upstream_fair）

fair

Allocate requests according to server response time , Priority allocation with short response time

url_hash

Press visit url Of hash Result allocation request , Make each url Directed allocation to the same back-end server , Multiple requests for the same resource , Requests may be made to different servers , Result in unnecessary multiple downloads , Cache hit rate is not high , And some waste of resources and time , Use url_hash, Can make the same URL Will reach the same server , When I ask again , Can be read from the cache .

- END -