[Tsvwg] Re: [e2e] Are you interested in TOEs and related issues

[Craig Partridge]

In message <200403081646.48030.zec at tel.fer.hr>, Marko Zec writes:

>On Sunday 07 March 2004 23:06, Alexandre L. Grojsgold wrote:

>> Small packets get more buffer frames, but are also more rapidly sent.
>> The gueue grows fats, but shrinks fast too. It means that a burst of
>> small packets will not increase the packet loss, no mather the length
>> of the buffer (in packets).
>
>
>This theory is unfortunately wrong, because it ignores the fact that 
>when congestions do occur, traffic flows consisting of small packets 
>(regardless whether they are bursty or smooth) will use more buffer 
>slots then the flows carrying the larger ones, leaving less remaining 
>buffer slots available for queuing further packets / bursts.

I must be missing something here.  I think Alexandre's generally right.
If the buffers load and empty according to the amount of data in them,
then packet size isn't an issue.

The only situations in which this isn't true, that I can think off the
top of my head, are:

    * If the router has places where it moves all the bytes in the packet
      buffer, regardless of how much data the buffer actually contains.
      Since this would cause extraordinary internal bandwidth challenges,
      I doubt anyone does it for any but the lowest speed routers.

    * If the router has a deep pipeline and the time at each stage in the
      pipeline is set to equal the average packet time rather than the
      minimum packet time.  But most folks carefully design router pipelines
      for minimum packet times precisely because of this risk.

So clue me in -- what detail of router design am I missing?

Craig