Message queuing overview

Message queue is the of basic data structure “ fifo ” A kind of data organization of . Think about it , Shopping in life , Need to line up , People who line up first buy and consume first , It's typical “ fifo ”.

 

 

 

image.png

MQ What problem to solve

MQ Is always there , However, with the popularity of microservice architecture , It has become a common tool to solve the problems between microservices .

 

The application of decoupling

Take e-commerce applications , There is an order system in the application 、 inventory system 、 Logistics system 、 Payment system . After the user creates the order , If the coupling calls the inventory system 、 Logistics system 、 Payment system , Any subsystem fails , All of these will cause abnormal order operation .

 

When it comes to message queuing , The problem of inter system calls will be reduced a lot , For example, the logistics system fails , It will take a few minutes to repair . In these few minutes , The memory to be processed by the logistics system is cached in the message queue , The user's order can be completed normally . When the logistics system is restored , Continue to process the order information , The single user cannot feel the failure of the logistics system . Improve system availability .

 

 

image.png

Flow peak elimination

Take a chestnut , If the order system can process up to 10000 orders , This processing capacity is more than enough to cope with orders in normal times , In normal time, we can return the result one second after placing an order . But at the peak , If there are 20000 orders, the operating system can't handle it , Only after the order exceeds 10000, users are not allowed to place an order .

 

Use message queue as buffer , We can lift this restriction , Divide orders placed in one second into a period of time to process , In this case, some users may not receive the successful operation of placing an order until more than ten seconds after placing an order , But it's better than the experience of not ordering .

 

Message delivery

Multiple service teams are interested in data , You only need to listen to the same kind of message to process .

 

image.png

for example A Generate data ,B Interested in data . If there is no queue of messages A After each processing, you need to call B service . After a while C Also sensitive to data ,A You need to change the code , call B service , call C service . As long as there is a need for service ,A All services need to change the code . Very inconvenient .

 

image.png

 

With message queuing ,A Just send the message once ,B Interested in news , Just listen for messages .C Interested in ,C Also listen to the news .A As a basic service, the service does not need to be changed at all .

 

Asynchronous messaging

image.png

Some calls between services are asynchronous , for example A call B,B It takes a long time to execute , however A Need to know B When can I finish , There used to be two ways ,A Call after a period of time B Query for api Inquire about . perhaps A Provide a callback api,B Call after execution api notice A service . Neither of these methods is very elegant

 

image.png

Use message bus , Can be very convenient to solve this problem ,A call B After service , Just monitor B Processing completed message , When B After processing , Will send a message to MQ,MQ Will forward this message to A service .

 

such A The service does not need to be called circularly B Query for api, No need to provide callback api. Again B Services don't have to do these operations .A The service can also get the message of successful asynchronous processing in time