Microservice architecture design pattern-(11) query

Two modes

• API composition model
  • Call multiple services to obtain data and then combine it into what you want
  • accomplish
    • API composition is done by the front end
      • The front end directly calls multiple services to query information
      • The front end is too heavy
    • API composition is done by the gateway
    • API composition as a standalone service
  • benefit
    • Simple and intuitive
  • Disadvantages
    • add extra overhead
    • Reduced availability
      • Because multiple services need to be called, and the availability of each service is 99.5%, then if the data of 4 services is to be aggregated, the service availability of the combiner becomes 97.5% (0.995^5, it is also a service in itself)
      • If you want to improve availability, turn some services intonon-blockingYes
        • If a service is unavailable, just ignore it and return incomplete data. In this way, other useful information can be displayed on the user page without directly reporting an error.
          • Of course, when this service is unavailable, maintenance personnel can be notified through monitoring alarms.
    • Lack of transactional data consistency
      • During the execution of the transaction, some parts of the query have been updated, and some have not yet been updated.
• CORS mode
  • Command query separation of duties
  • Use events to maintain read-only views copied from multiple databases for querying
  • benefit
    • Efficient
      • No need to consume memory and network like API combination
    • Different types of query implementations
      • For example, there is a service that uses a NoSQL database and cannot support queries very well. Then using CQRS to build a dedicated query database at this time can solve this problem very well.
        • Another example is parsing MySQL's json data into fields that query the database.
    • Event source query
      • The information stored in the event is placed in the json of event_data, which is not conducive to querying. Functional query event ID or aggregate ID. Use CQRS to construct one or more latest views, which can be queried very well
    • isolation
      • Isolate commands and queries
  • Disadvantages
    • complex architecture
    • Need to deal with delays caused by data replication
      • How to solve
        • Let the client record the latest event ID tag
        • Then bring it with you when querying. If the query view has not been updated and the ID cannot be found, an error will be reported.
        • Because the function ensures that the data is ultimately consistent.