With the development of the company where the author is located, the scale of application services has been expanding, and the original vertical application architecture has been unable to meet the development of the product. Dozens of engineers have developed different functions in parallel in one project, and the development efficiency has been continuously reduced, so the company started. Fully promote the service process, and concentrate most of the engineers' energy in the team into microservice.

Microservice allows each engineer to develop only in the subprojects that he is responsible for, improving the efficiency of development, but the service also brings other problems:

• It is impossible to know the operation of each service, for example, what is the current QPS of a service? What is the average delay, the delay of 99%, and the delay of 99.9%?
• The response time of an api is slow. How to locate which method is caused?
• Is the load of each service balanced?
• When the service is jittery, how do you determine whether it is caused by the DB or caused by the Cache?
• What is the Cache RT?
• How to assess the capacity of the service, as the number of calls to the service increases, how many machines does the service need to support? When should I add a machine?