Method for realizing full distribution of protocol stack process and distributed system

Abstract

The invention discloses a method and a system for realizing full distribution of a protocol stack. In the method, the full distribution aim of the protocol stack is fulfilled by synchronizing a small number of internet protocol control block (INPCB) data according to the characteristics of a transmission control protocol (TCP), a user datagram protocol (UDP) and a raw internet protocol (RawIP). In particular, for the TCP, the INPCB data of a mother SOCKET is synchronized; for the UDP, the INPCB data connected with a server terminal is synchronized; and for the RawIP, all the connected INPCB data is synchronized. By the method and the system, the full distribution of protocol stack process can be realized without synchronizing all the INPCB data.

Description

technical field [0001] The invention relates to a processing technology of a distributed system, in particular to a fully distributed implementation method of protocol stack processing in a distributed system and a distributed system. Background technique [0002] At present, in order to improve the processing capability, a distributed system has appeared. The distributed system includes multiple subsystems. Each subsystem has an independent CPU processing capability and can run the protocol stack independently. Multiple subsystems cooperate to complete a set of system functions. [0003] In a distributed system, it is necessary to support the distributed processing of the protocol stack. The protocol stack refers to the Transmission Control Protocol / Internet Protocol (TCP / IP) protocol stack, including the three INET protocols of the Transmission Control Protocol (TCP), User Datagram Protocol (UDP), and Raw Internet Protocol (RawIP). In the INET protocol family (including I...

AI Technical Summary

Problems solved by technology

[0011] 3. For the RawIP protocol, because there is no port feature, it cannot participate in the division of the port range. Such a connection is limited to the establishment of the main subsystem. After the subsystem receives the RawIP connection message, it sends it to the System Master for centralized processing.

[0016] Due to the restriction that only the System Master can be used as the server to listen to well-known ports, and the port-based distribution can only support protocols such as TCP / UDP that need to be identified by port numbers. For protocols without port information such as RawIP, only in System Centralized processing on the Master, so in fact only partially distributed

The consequence of realizing partial distribution is to increase the burden of System Master, thereby reducing the performance of System Master

[0017] 2. Vulnerable to packet attacks

When the user constructs a large number of well-known port messages to enter from each subsystem, each subsystem checks that the destination port is a well-known port and sends them all to the System Master for processing. However, the corresponding services may not be enabled on the System Master, which increases the burden on the System Master. processing, affecting normal business processing

[0019] 3. The scalability of the system is not high

[0020] When each subsystem acts as a client, it can only allocate local port numbers from the pre-agreed non-well-known port range, and cannot be expanded. Therefore, the local dynamic port resources that can be used by each subsystem to establish a connection are limited, and it is easy to form a system bottleneck.

Moreover, port resources cannot be fully utilized. For example, if a subsystem does not establish any connection, the port resources agreed to this subsystem cannot be used by other subsystems, which wastes port resources.

Images