Master-slave architecture server, service processing method thereof and service processing system thereof

Abstract

The invention discloses a master-slave architecture server, a service processing method thereof and a service processing system thereof. The service processing method of the master-slave architecture server comprises the steps of analyzing and uploading a configuration file by a parent process of the master-slave architecture server, wherein the configuration file includes multiple ports and a process serial number which is in one-to-one correspondence with each port; distributing different process serial numbers for established sub-processes by the parent process; inheriting corresponding resources from the parent process based on the process serial number by each sub-process, wherein the resources include ports; and, by the sub-processes, monitoring corresponding ports to receive a service request from a client, processing the service request and then returning a service response to the client. According to the embodiment of the invention, as different sub-processes respectively monitor reading-writing events of different ports, the service request of one port can be completely processed by one sub-process, the sub-processes are completely independent and no longer compete with each other for the service request, and a thundering-herd phenomenon is accordingly avoided.

Description

technical field [0001] The present application relates to the field of computer technology, and in particular to a master-slave architecture server, a method for processing business, and a system for processing business. Background technique [0002] Currently, many servers have a multi-core central processing unit (CPU) and support a Symmetric Multi-Processing (Symmetric Multi-Processing, SMP) architecture. In order to make full use of CPU resources, many high-performance servers adopt a master-slave architecture. The schematic diagram of the master-slave architecture is shown in figure 1 As shown, the parent process first parses and loads the configuration file, creates a socket and binds the listening port, and then creates multiple child processes, each of which inherits the resources of the parent process. The number of sub-processes is the same as the number of cores of the CPU. In order to prevent sub-processes from migrating between CPUs, sub-processes can also be b...

AI Technical Summary

Problems solved by technology

[0005] (1) Multiple working processes listen to the same port at the same time. When a request is sent to a certain port, multiple working processes will compete for the right to process the request. In the end, only one process can successfully compete, which causes a shocking crowd Phenomenon;

[0006] (2) In a real production environment, especially in a virtual machine environment, some processes are more likely to succeed in competing requests than other processes, causing some processes to receive a large number of requests and be overly busy, while other processes cannot get requests and are overly idle. The load imbalance between processes cannot give full play to the advantages of the SMP architecture;

[0007] (3) When the working process exists, that is, some requests can only be processed correctly if they are received by a specific working process. In this case, if these requests are received by other processes, the process receiving the request can only seek help from other processes , which increases the communication cost

For example, each working process independently maintains its own cache, which stores the user's personal information. If the personal information of user A is in the cache of process B, when the request initiated by A is received by process C, since process C does not have user A personal information, so it is impossible to directly give a correct answer to user A

At this time, it can only be obtained from process B through process communication. This behavior not only increases the execution complexity of processes B and C, but also increases the response time to A's request;

[0008] (4) The configuration of the working process is uniformly inherited from the parent process, and the configuration is the same and cannot be fine-tuned;

[0009] (5) In order to solve problems such as thundering herd phenomenon, load imbalance and unnecessary inter-process communication, it is inevitable to introduce process locks. Multiple processes solve the synchronization and mutual exclusion problems between processes through process locks, but the introduction of process locks It will also increase the execution complexity of the process and reduce the performance of the server

Images