High concurrency ：1. Improve the performance of a single machine or web Performance of ;2. Deploy multiple services to support .

High availability ：1. redundancy ;2. Fail over ( perception , monitor )

                  Technical means ： colony （redis Sentinel mode ;）;keepalived Survival detection

Nginx Reverse agent horizontal expansion & High availability analysis

If nginx Not enough use , Sure DNS The same domain name is configured in load balancing mode to resolve multiple IP, Theoretically unlimited expansion .

  High availability analysis ：DNS The server and Nginx Basically, there are two services , Failure awareness is not possible . If so configured, then 01 Whether the service is good or not will be parsed and returned . To be highly available, you have to give nginx1 Redundant one nginx1_1, But what is on display IP All are 1.0.0.1. Technology is using VirtualIP Virtual one IP, Use KeepAlived Probe nginx Usability , If nginx1 Hang up , Automatic access nginx1_1.

Gateway horizontal expansion & High availability analysis

Can be in nginx.conf More than one server, And configure load balancing rules .nginx Can detect the back server Usability , Theoretically, infinity can be added server, But limited to nginx Throughput .

High availability analysis ：nginx Can detect tomcat Whether it works well , If it doesn't work well, you won't forward the request to this service .

Service level expansion & High availability analysis

Call between services , It's all through RPC Way to call , To decouple the caller and the backcaller , Separate from the configuration center , Both the called and the called are registered in the configuration center, such as Zookeeper,Eureka,Nacos etc. , Configuration center

High availability analysis ： Between services RCP Call to client Rotation training visit ProductIP Address , If the access is different, visit the next Server, And there is a special background thread to synchronize with the registry ProducerServerIP list .

High concurrency of data cache & High availability analysis

The most widely used cache is redis,mongodb,rockermq Components , Components use their own clusters and safeguards .

Database horizontal expansion & High availability analysis

DB The expansion is divided into the following centralized ：

1. The business will split the relatively cohesive data into another library
   1. It's a way to split the architecture , Can't dynamically expand
2. Order data and historical order data within three months In different libraries
   1. Hot and cold data requires migration
   2. Theoretically, the order volume will exceed one database capacity within three months , This method will fail
3. Route to different databases according to the location of the order
   1. It's the second deformation above
   2. No data migration required , But the data of a region in Beijing will also exceed , This is not easy to use
4. Just according to ID Conduct Hash Calculation , Try to distribute them equally among different DB
   1. Not very feasible , There is no problem storing data , Visit ？ You don't know the data in that database , So this kind of unreality
5. According to the read-write attribute, it is divided into read library and write library
   1. Write to the main database , Reading data from the library
   2. If reading concurrency is high , You can have several Slave Support
   3. For more than salve Read from library , If a slave library goes down ,druid The connection pool will detect and obtain other connections for access
   4. Write library There is a main library , This requires redundancy .

High availability analysis ： What time is it Mysql It does not meet the requirements of high availability , Need a redundant machine . similar nginx High availability , Use virtualIP+KeepAlived To detect and failover , I need to pay attention to my two Mysql The server may have poor data synchronization .

END