|
Posted by Patrick Klos on March 23, 2007, 10:21 am
If you were Registered and logged in, you could reply and use other advanced thread options
>
>Hi,
>
>I have a question about ethernet switches on full duplex "collision-
>free" LANs.
>
>What if there are multiple nodes sending to the same segment/port, how
>does the switch handle that? I would think it could only let through
>one of the frames.
The switch actually has buffers to queue up the packets for a given
port. When more than 1 sender tries to send packets to the same receiver,
the switch buffers the packets while the port is busy. Of course, if too
many packets are sent to the same port, the buffers could be exhausted and
packets will be lost.
>I guess that's not really considered a "collision" since one does get
>through. Is that right?
No, that wouldn't be considered a collision.
>How do the other nodes recover from that? In full-duplex mode does
>Ethernet just not do any resends and let higher layers worry about it?
Except at the lowest level, ethernet always lets the higher layers worry
about resending lost packets.
>It seems like the switch could just tell a little fib and say to the
>nodes that sent the ones that didn't get through that a collision
>occured, so Ethernet could handle the resend. Maybe that doesn't make
>sense, but otherwise it just seems like you would have to wait for
>destination node to tell you that a packet has been lost and that
>would take more time.
There are some flow control mechanisms for ethernet, but I am not too
familiar with their implementation.
>I was just wondering about how this actually works, and I would
>appreciate it if some one could clear this up, I would appreciate it.
The basic answer to your question is that switches have buffers to queue
packets up until it can send them out their target port(s). Hope that
helped?
Patrick
========= For LAN/WAN Protocol Analysis, check out PacketView Pro! =========
Patrick Klos Email: patrick@klos.com
Klos Technologies, Inc. Web: http://www.klos.com/ ============================================================================
|