A method and system for receiving network driver layer data packets based on dynamic prediction

Abstract

The invention discloses a dynamic prediction-based network driver layer data packet receiving method, which comprises the following steps of establishing a blank chain table and a circulation sequence, and creating a kernel thread and a kernel timer; receiving network equipment interruption, executing an interruption processing program, and stopping receiving network equipment interruption immediately after entering the interruption processing program; judging whether a network equipment interruption instruction is a receiving instruction or a wrong instruction; if the instruction is a receiving instruction, judging whether a data arriving bit in a receiving descriptor of the receiving instruction is zero or not, if not, judging whether the number of data packet caches in the blank chain table is lower than a threshold or not, if not, obtaining a new data packet cache from the blank chain table and judging whether obtaining is successful or not, if not, counting the length of effective data in the data packet cache corresponding to the current receiving descriptor, and discarding the effective data. By the method, the number of the pre-distributed data packet caches can be calculated to adapt to the change of the network flow, and the phenomenon of consuming too much memory when the data flow is lower is avoided.

Description

technical field [0001] The invention belongs to the field of network technology, and more specifically relates to a method and system for receiving network driver layer data packets based on dynamic prediction. Background technique [0002] The network driver adopts two methods in the process of receiving data packets, the traditional interrupt I / O method and the NAPI (New API) method. [0003] The traditional interrupt I / O method only processes one data packet during each network device interruption process. The advantage is that when the data traffic is not very large, the response is timely and the system resources are saved; Interrupt calls will occupy too much CPU time and affect the performance of the server. [0004] NAPI (NEW API) is a data packet receiving method adopted in the Linux system that is different from the traditional method. The NAPI method combines interrupt triggering and active polling to receive data, reducing the frequency of data packet receiving ...

AI Technical Summary

Problems solved by technology

[0003] The traditional interrupt I / O method only processes one data packet during each network device interruption process. The advantage is that when the data traffic is not very large, the response is timely and the system resources are saved; Interrupt calls will take up too much CPU time and affect server performance

[0004] NAPI (NEW API) is a data packet receiving method adopted in the Linux system that is different from the traditional method. The NAPI method combines interrupt triggering and active polling to receive data, reducing the frequency of data packet receiving interruptions and high-frequency interrupt scheduling. System overhead; it has high efficiency when the data flow is high and the data packet is small; but it still has some disadvantages: due to the software interrupt method, there will be a certain delay. When the data flow is low, the NAPI method is more The traditional interrupt I / O method is less efficient, and when the data packet is large, applying for data packet cache and copying data will take a lot of processing time

Images