互联网分片和重组程序
internet fragmentation and reassembly procedure
我目前正在阅读 RFC 791 of the IETF which explains the Internet Protocol thoroughly. At one place it talks about fragmentation and reassembly procedure based on network types. On researching further I came across a research paper 基于 MTU(最大传输单元)的分段概念。我有两个问题。首先,分段和重组过程是否代价高昂(我希望它们是因为有新数据报的创建和大量复制),如果是的话,其次,路由协议是否考虑了具有更大 RTT 的更长路径但也有更大的 MTU,这样如果数据报通过它们,则需要更少或不需要碎片?
Firstly, are fragmentation and reassembly costly procedures (which I
hope they are as there are the creation of new datagrams and a lot of
copying)
这取决于您的想法"costly"。但一般来说,是的。碎片化是您要避免的事情。既浪费又危险(丢一个分片,就丢整个包)。
secondly, do the routing protocols take in consideration of a longer
path with a greater RTT but also having a greater MTU such that less
or no fragmentation is required if the datagrams proceed through them?
没有。路由器不知道不同目的地的 MTU。实际上,网络通常使用 1500 字节的恒定 MTU,尤其是核心网络。据我所知,在实践中您可能 运行 进入碎片的唯一地方是由于隧道而在网络边缘。
我目前正在阅读 RFC 791 of the IETF which explains the Internet Protocol thoroughly. At one place it talks about fragmentation and reassembly procedure based on network types. On researching further I came across a research paper 基于 MTU(最大传输单元)的分段概念。我有两个问题。首先,分段和重组过程是否代价高昂(我希望它们是因为有新数据报的创建和大量复制),如果是的话,其次,路由协议是否考虑了具有更大 RTT 的更长路径但也有更大的 MTU,这样如果数据报通过它们,则需要更少或不需要碎片?
Firstly, are fragmentation and reassembly costly procedures (which I hope they are as there are the creation of new datagrams and a lot of copying)
这取决于您的想法"costly"。但一般来说,是的。碎片化是您要避免的事情。既浪费又危险(丢一个分片,就丢整个包)。
secondly, do the routing protocols take in consideration of a longer path with a greater RTT but also having a greater MTU such that less or no fragmentation is required if the datagrams proceed through them?
没有。路由器不知道不同目的地的 MTU。实际上,网络通常使用 1500 字节的恒定 MTU,尤其是核心网络。据我所知,在实践中您可能 运行 进入碎片的唯一地方是由于隧道而在网络边缘。