RPC protocol details

One 、 What is? RPC agreement

Remote procedure call protocol , It is a way of requesting services over a network from a remote computer program , You don't need to know the protocols of the underlying network technology .
The specific calling process is :
1. Serving consumers （client client ） Call the service that needs to be consumed by calling the local service ;
2. Client stub （client stub） After receiving the call request, it is responsible for putting the method 、 Serialization of input parameter information （ assemble ） To be a message body capable of network transmission ;
3. Client stub （client stub） Find the remote service address , And send the message to the server through the network ;
4. Server stub （server stub） Decode the message upon receipt （ Deserialization operation ）;
5. Server stub （server stub） Call the local service according to the decoding result for relevant processing ;
6. The local service executes the specific business logic and returns the processing result to the server stub （server stub）;
7. Server stub （server stub） Repackage the returned results into messages （ serialize ） And send it to the consumer through the Internet ;
8. Client stub （client stub） Message received , And decode it （ Deserialization ）;
9. Service consumers get the end result ;

and RPC The goal of the framework is to achieve the 2-9 The steps are well sealed , That is to call 、 code / The decoding process is encapsulated , Let the user feel like calling a local service to call a remote service .

Two 、 Realization RPC Problems to be solved

Communication problems
The remote invocation , There must be communication between the two machines . Mainly through the establishment between the client and the server TCP Connect , All the data exchanged by the remote procedure call is transferred in this connection . Connections can be on-demand , It will be broken after the call , It can also be a long connection , Multiple remote procedure calls share the same connection .
Usually this connection can be on-demand （ When it needs to be called, establish the connection first , It will be broken immediately after the call ）, It can also be a long connection （ Keep the client and server connected for a long time , No matter whether there is a packet sent at this time , It can cooperate with heartbeat detection mechanism to regularly detect whether the established connection is alive and effective ）, Multiple remote procedure calls share the same connection .

Service addressing
above-mentioned ,RPC It is local oriented . Then the user makes a local call , How to find the corresponding remote call , need RPC Implement at the bottom . Need to manage ip, port , Service list and other information . It's common to use zookeeper To manage .
For producers :
When the service starts , Register your services to the Registration Center ;
When the producer hangs up , The registry logs out of the stopped service ;
The producer sends heartbeat to the registration center regularly , If the heartbeat is abnormal , The registry also unregisters producer services .

For consumers :
Subscribe to the service of the registry , obtain ip, Port and other service information ;
When producers go online and offline , The registry will notify the caller ;
After the service is offline , Consumers no longer consume the service .

Network transmission
When A The application on the machine initiates a RPC Invocation time , Call method and its input parameter information need to go through the underlying network protocol, such as TCP Transferred to the B machine , Because the network protocol is based on binary , All the parameter data we transmit needs to be serialized first （Serialize） Or group （marshal） Only in binary form can it be transmitted in the network . Then the binary data after serialization or marshalling is sent to B machine .

When B The machine received A After the request from the application of the machine , It also needs to deserialize the received parameters and other information （ The reverse operation of serialization ）, To restore binary information to an in memory representation , Then find the corresponding method （ Part of addressing ） Make a local call （ It's usually through the generation of agents Proxy To call , Usually there will be JDK A dynamic proxy 、CGLIB A dynamic proxy 、Javassist Bytecode generation technology, etc ）, Then get the return value of the call .

The service call
B The machine makes local calls （ Through agency Proxy And reflection calls ） Then we get the return value , At this time, you need to send the return value back to A machine , It also needs to be serialized , And then send the binary data back to A machine , And when A After the machine receives these return values , Then deserialize again , Revert to in memory expression , Give it to A The application on the machine carries out relevant processing , Generally, business logic processing operations .

3、 ... and 、RPC And REST The difference between ？

RPC And REST The biggest difference is RPC Provides better abstraction ,RPC Even completely hide the details of network transmission , and REST No, . say concretely ,REST As a minimum, it is required to provide URL And request parameters , and RPC Hide the implementation details related to network transmission . On the other hand ,RPC It can be based on any network communication protocol , and REST Usually based on HTTP（ perhaps HTTPS） agreement .RPC The caller doesn't care if the specific agreement is ：HTTP、TCP Or any other custom protocol .

Four 、 Advantages and disadvantages

advantage :
Help the client communicate with the server through the procedure call of traditional high-level language .
It can be used in distributed environment and local environment .
Support process oriented and thread oriented models .
Hide the internal messaging mechanism from users .
Code can be rewritten and redeveloped with minimal effort .
Provide abstraction , That is to hide the message transmission nature of network communication from users .
Omit many protocol layers to improve performance .

shortcoming :
The client and server use different execution environments for their routines , And resources （ for example , file ） The use of is also more complex . therefore ,RPC Systems are not always suitable for transmitting large amounts of data .

RPC Very prone to failure , Because it involves communication systems , Another computer and another process .

RPC There is no uniform standard ; It can be realized in many ways .

RPC Just based on Interaction , therefore , In terms of hardware architecture , It does not provide any flexibility .

Reference article :
RPC Protocol analysis [ Easy to understand ]
Thousand words take you to know what is RPC agreement