PCBUILD Archives

Personal Computer Hardware discussion List

PCBUILD@LISTSERV.ICORS.ORG

Options: Use Forum View

Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Dave Gillett <[log in to unmask]>
Reply To:
PCBUILD - Personal Computer Hardware discussion List <[log in to unmask]>
Date:
Sun, 30 Jan 2000 18:18:34 -0800
Content-Type:
text/plain
Parts/Attachments:
text/plain (60 lines)
On 29 Jan 00, at 13:14, Herbert Graf wrote:

> > Also, are you running CAT5 cable, BNC, or what?  Another cause of
> > packet collision can be interference such as EMF (electro magnetic
> > frequency) or RF (radio frequency).  Are the cables paralleling any power
> > lines or do you have any other devices near the hub that might be putting
> > out these kinds of signals.  Keep in mind that these would have
> > to be pretty
> > strong signals if you are running shielded cable for them to cause any
> > interference.
>
>         Running 10BASE-T, 5 port hub is right in front of me
> (that's how I can see the collision light, it's like an external
> modem, I love seeing the blinking lights!). I tried moving it, which
> didn't help, but since the problem only happens on the start of a
> transfer I don't believe that is the cause.



  This is Ethernet with a hub.  That means that it works a bit like
CB radio (remember those?).  Everybody on the network *can* hear
everybody else, and part of each NIC's job is to pick out only the
packets that are intended for it.

  When a NIC has a packet to send, it listens for a moment to see if
the "air" ("Ether"...) is clear.  If it's not, it waits.
  When it finds that the air is clear, it begins sending.  However,
there's a chance that some other NIC listened to the same clear
space, and started sending at almost exactly the same time.
  THIS IS A COLLISION.  Not big packet sizes, not eletro-magnetic
interference, and not -- so far -- anything that can't happen on a
*perfectly healthy* Ethernet network.

  Another part of the NIC's job is to listen to the net while it
sends, and detect any collision.  If a collision occurs, the packet
cannot be received at its destination, and so the NIC will pause and
then try again.  The length of the pause is randomized so that two
NICs, having collided once, will not collide again when they attempt
to retransmit.
  However, a retransmission attempt *can* collide with some third NIC
attempting to transmit, and so cause further retransmissions.  So
while a smallish number of collisions are normal and healthy, a
network where collisions are happening constantly has difficulty ever
moving any packets.  This scenario is why it's useful to have an LED
that monitors collisions.

  Note that collisions can only occur in the first portion of any
packet, since the colliding NIC must have "heard" clear air just a
moment before trying to transmit.  That, I believe, is why you tend
to see it "on the strat of a transfer".

  [It is possible to eliminate collisions, by replacing the hub with
a switch.]

David G

               PCBUILD mailing list is brought to you by:
                            The NOSPIN Group
                  http://nospin.com - http://nospin.org

ATOM RSS1 RSS2