Cable Forum - View Single Post

Ignition · 13-02-2005, 15:57

Pretty please do a Google on how routers handle standard traffic and how they handle pings to themselves, there are completely different logical paths - if any more questions post.

Actually no forget that quick explanation.

Routers have at least 2 different processing components.

A generic CPU for doing the mundane stuff, calculating routes, taking in and processing commands, responding to ICMP traffic and packets with a destination address that's one of the router's interfaces, etc, etc.

The other part is usually a dedicated Application Specific Integrated Circuit (ASIC) based Switching/Routing/Forwarding engine which takes packets in, strips the headers and decides based on a routing and a forwarding table (maintained through interaction with the CPU for ARPing, retrieving information from routing protocols, etc) which of its' interfaces to shove the packet out of, and then re-encapsulates the packet in new headers.

Most packets the ASIC based engine knows what to do with them due to previous packets and the CPU is not involved in any way. Only when dealing with a completely new network is CPU interaction required.

I hope this helps a bit. Also remember routers don't give a monkeys about whether the traffic is TCP or UDP, they work on IP level and won't look upwards of this unless told to, which requires CPU as the routing engine only deals with IP and below.

13-02-2005, 15:57	#29
Ignition Permanently Banned Join Date: Jun 2003 Location: South-East London Age: 48 Services: Depends who's being serviced :p Posts: 2,588	Re: Packet Loss ?!? Pretty please do a Google on how routers handle standard traffic and how they handle pings to themselves, there are completely different logical paths - if any more questions post. Actually no forget that quick explanation. Routers have at least 2 different processing components. A generic CPU for doing the mundane stuff, calculating routes, taking in and processing commands, responding to ICMP traffic and packets with a destination address that's one of the router's interfaces, etc, etc. The other part is usually a dedicated Application Specific Integrated Circuit (ASIC) based Switching/Routing/Forwarding engine which takes packets in, strips the headers and decides based on a routing and a forwarding table (maintained through interaction with the CPU for ARPing, retrieving information from routing protocols, etc) which of its' interfaces to shove the packet out of, and then re-encapsulates the packet in new headers. Most packets the ASIC based engine knows what to do with them due to previous packets and the CPU is not involved in any way. Only when dealing with a completely new network is CPU interaction required. I hope this helps a bit. Also remember routers don't give a monkeys about whether the traffic is TCP or UDP, they work on IP level and won't look upwards of this unless told to, which requires CPU as the routing engine only deals with IP and below.