Interviewer: how to deal with high concurrency?

I chatted with my high school classmates yesterday , Large factories need high concurrency experience and cannot enter , Only enter small workshops , There is no need to deal with high concurrency in small workshops , Such an endless cycle . High concurrency experience , It will never be out of date in the interview scene , It plays an important role , Spike scenario traffic peak shaving is one of the high concurrency .

What is high concurrency ？

High and pointed at the same time , A large number of customer requests , Access server ,update Modify database data , Now update Lock the watch , Wait for execution to complete before processing the next request , When customer requests accumulate to a certain amount , Database link limit exceeded , Will return link timeout , Also because of too many requests , Add the same data twice , There is no guarantee of data consistency . The core element of high concurrent requests is to ensure that a large number of requests are highly available , Uniformity , High performance . As can be seen from the above , Ensure the stability of the system , It can reduce the direct impact on DB Request update of layer data .

How to ensure high performance ？

The customer layer can adopt the methods of user answering questions and entering verification code , Limit the number of requests , prevent 0 Click the moment of seconds , The peaks are all concentrated , The most important thing is to prevent someone from cheating and killing , And it can also delay customer requests , Bar requests are distributed in different time periods .

The business layer can limit the number of interface calls .

Flow peak shaving can also be used to achieve , Nowadays, there are a large number of excellent open source frameworks supporting traffic peak shaving , Such as RabbitMq and kafka etc. .RabbitMq In addition to the functions of decoupling and asynchrony , It can also realize flow peak shaving , Store a large number of user requests in mq In line , Then you can allocate and enter the database in other time periods .

How to ensure consistency ？

When consuming warehouse inventory , Before each consumption , Judge once count>0, When there is inventory , Can we continue to consume and execute , Otherwise, roll back .

How to achieve high availability ？

Last , In order to ensure the high availability of the server , It must be deployed on multiple servers to ensure that one of them is down , The system is still working .

service degradation

When the request reaches the system capacity , You can turn off other non core functions first , Try to give all the system memory to the seckill function , Ensure the normal operation of secsha .

Service restriction

Flow restriction refers to the flow restriction of the access request of the interface , It can also be used. rabbitmq Carry out current limiting ,mq The queue flow will be limited , Once the expiration time is exceeded , Will be discarded , The client failed quickly and tried again in the second round .

Refuse request

When service degradation and service flow restriction cannot be solved , Finally, only one error page can be returned to the user , such as “ Too many user requests ”“ Service is busy , Try again later ” etc. , It only happens when the server is overloaded , There will only be temporary unavailability , Once the normal server load drops , You can normally request .