tilting at windmills discussion 2

Wed Jun 19 15:44:57 UTC 1996

> In a similar vein, something to think about:
> 
> IP packet size distribution (970780293  total packets):
>    1-32   64   96  128  160  192  224  256  288  320  352  384  416  448  480
>    .000 .497 .094 .013 .009 .006 .007 .010 .010 .014 .004 .005 .002 .002 .002
> 	^^^^
>     512  544  576 1024 1536 2048 2560 3072 3584 4096 4608
>    .004 .002 .189 .000 .069 .053 .000 .000 .000 .000 .000
> 
> Collected at a private interconnect between CICNet, OARnet and OSU. over a
> period of a week or so.

It would be really interesting if you could expand the range of 32 to 64
byte packets to display in one byte increments.  Then we could
try to guess:

o	How many packets are nothing but TCP and IP headers?

o	How many packets appear to be TCP/IP headers with one byte of
	data?

o	How many of these packets, with some sort of header compression
	scheme, could fit in one ATM cell?

o	Etc., etc.

-tjs