Re: WG ACTION: 2 weeks to discuss [LL62] Do not space probes randomly

Philip Nye wrote:

>the chance we need to avoid is that you and I 
>both agree on ALL FOUR values in our series.

This is statement is asserted without any supporting evidence, and it is 
false.

This is NOT what we need to avoid.

In order to be a good network citizen, we want to minimize the aggregate 
peak load that a large number of hosts may impose on network 
infrastructure, since peak load is what causes loss. This issue only 
arises when there are a large number of hosts probing simultaneously. Two 
hosts, or three, or four or even ten host probing at *exactly* the same 
time is NOT a problem.

The Ethernet collision loss myth has existed for decades. University 
computer science lecturers teach students that it's a myth, but still the 
myth persists in the larger population.

Collisions do not cause loss on Ethernet. They did not on thick coax, 
thin coax, or on UTP.

Pay particular attention to the second paragraph below.

<http://www.postel.org/pipermail/end2end-interest/2001-October/001529.html>


>What Vernon says is quite right and I'll only add that Collision sensing
>and recovery happens in times on the order of 200 uS to a few mS, on
>even extremely highly contended Ethernet segments (many stations ready
>with pkts to send all the time).  This is far faster than Token passing
>(many mS), in any of its forms, as a very pertinent graph from the
>original IEEE 802 standarization simulations shows (it can be faxed to
>anyone who wishes a copy).  This graph is particularly telling in that
>CSMA/CD become better in relation to Token as the number of contending
>nodes increases -- yes, better.
>
>One reason for trying to rid the world of 'collision' myths is that they
>serve as an alarm for how easily misinformation can arise and how hard
>it is to gather peoples' intellects together to stamp the myths out. 
>Confining ourselves to networking, we must be aware that this applies in
>the TCP/IP realm as well.  Back to the LAN realm, which the original IP
>world had no cognizance of, the collision myth, that started as an IBM
>marketing weapon, matured into a switch vendors' scare tactic to sell at
>first cut-through (a bad idea now in disrepute) and later "a segment for
>every node", so that "collisions would no longer occur" -- just pushing
>the problem into provisioning of the switch fabric.  "Caveat emptor" is
>as necessary today as it has ever been.

Ralph Droms wrote:

>If there is no randomization of subsequent probes, there is no point in
>specifying multiple probes.

No, this is false.

We don't assume the ARP probes are the ONLY traffic on the network.

The reason for trying three times is that there may be OTHER brief bursts 
of traffic on the link that overflow the switch buffers (or cause CSMA/CD 
breakdown, etc.) resulting in one or more probes being lost.

In reality, the amount of traffic generated by ARP probing is too low by 
itself to cause switch buffer overflow -- PROVIDING that it is 
distributed over several seconds. Should all the probes be sent in a 
tight bunch, then the chance of switch buffer overflow could become 
non-zero, which is why the initial random delay is useful to eliminate 
that.

Stuart Cheshire <[email protected]>
 * Wizard Without Portfolio, Apple Computer, Inc.
 * www.stuartcheshire.org