分片与多播基于网络处理器的高速实现

High-speed Implementation Fragment and Multicast Based on Network Processor

传统的IP数据报文的分片和多播技术,由于在处理过程中需要对整个报文内容进行多次数据拷贝,由此会对缓存报文的处理器外部存储空间进行大量I/O操作,这将严重制约了当前处理器,尤其是网络处理器(NP)的高速报文处理能力。文中介绍了一种新的分片和多播的实现方法,它针对于网络处理器自身的结构特点,对传统方法做出了一定的改进,无需进行大量数据复制和拷贝,即可完成一份数据报文的分片和多播处理,更适用于当今的高速网络交换设备。从而保证更高速的数据处理交换得以实现,同时,实现方法简单,可在任意软件环境下实现。

Traditional IP datagram fragmentation and multicast technology have restrained from data process ability of the processor (network processor (NP) especially), because it needs to copy lots of data many times of the datagram and results in a large amount of I/O operations to the buffer memory. This text introduces one new implementation method of the fragmentation and multicast, it aims at the network processor's characteristics, makes sure improvement to the traditional method, do not need the copying of lots of data, and can immediately complete datagram fragmentation and multicast function, more suitable for the high-speed network equipment nowadays. It guarantees more high-speed data exchanging, and it is simple to realize, can be finished under any software environment.