Re: calculate saving ratio of *link bandwidth*

I just did this on my box at home

1) netstat -n -i -b -I ppp0

        to measure the bytes in and out on my ppp link.

I observed it was quiescent, and fetched a page.

the page was http://mirror.aarnet.edu.au which is a table of text with
three small logos. I disabled graphics.

The size of the ppp level bytes transferred was 9181.
The size of the file saved to disk was 8518

so the overhead from TCP and IP (but I think excluding ppp framing) was 663
bytes, safely between 5 and 10% of the size of the html markup

THIS IS A SHORT PAGE so the relative % of HTML/HTTP header to content has to
be higher than for large binary data doesn't it?

This also excludes DNS: I had the page cached. DNS is 53 byte packets isn't
it?

I was on a modem. That makes my MTU 1500 at most, more like 1200 so I probably
fragmented, and again had more overhead than on 'real' media.

I agree you may well see more than this, but it seems to me that HTTP/HTML
as data in itself, has around 5% to 10% of overhead to be on the wire as IP
and I really don't see where the excess comes from to double that load, in
order to do a cache fetch.

Can this maybe be an artifact of 'transparency' and its effect on the
interface counts?

-George
Received on Fri Jun 12 1998 - 06:09:55 MDT