Frank O'Connor foconnor at ozemail.com.au
Tue Apr 7 23:45:12 AEST 2009

I tend to have those problems on the Net even now ...

Technically I have a connection that rates at between 15Mbs (on a 
good day) and 9Mbs (on a bad day).

Most sites on the Web don't funnel data at those speeds. FTP 
transfers rarely hit anywhere near my maximum and more usually at 
about 25% of my maximum. Peer to peer is good if it hits one third of 
my connection speed. Streaming and the like tends to get about 1/3 of 
my connection speed ... and I'm talking quick local sites like the 
ABC and that here.

The bottom line is that the Net is as fast as the slowest 
server/component/node in your current connection, and rarely meets 
optimum levels that the ISP's advertise. (And my connection speed 
does meet the advertised rate of the ISP.)

And don't get me started about site and server latency in the current 
environment - especially those sites that hook advertisers into their 
Web pages. Even with the current latency of PON enabled fibre, it 
couldn't be worse than many of these sites.

I suppose the point is that even without the technical limitations 
mentioned (and I'm sure that over 8 years many of these would be 
surmounted ... hey, progress happens!) there are any number of 
dysfunctional attributes (server performance, site performance, 
application performance, third party performance, router performance, 
firewall performance, etc etc) on our Net that are unlikely to be 
remedied by fixing the backbone and the hardware.

Yes, we are a long way from Utopia folks ... but I expect the 
hardware to be fixed long before the software.


At 11:09 PM +1000 7/4/09, Paul Brooks wrote:
>George Bray wrote:
>>  So what *are* the capacity/distance properties of today's FTTP fibre
>>  systems?  Paul? Richard?
>For *tomorrows* FTTP systems, there is labwork being done on WDM-PON -
>where the fibre is still split 32 ways up to 64 ways in the middle of
>the physical cable network, however in some implementations every
>endpoint gets a dedicated wavelength, with the 32 or 64 wavelengths
>being sent from the head-end.
>This effectively takes out all the sharing aspects - every customer gets
>a dedicated gigabit or even 10Gig optical path direct from the head-end
>to their premise! (drool).
>Now try to imagine the backhaul required to service the head-end if
>5000-10000 users are hanging off the head-end with 10Gig dedicated links
>each ;-)
>Link mailing list
>Link at mailman.anu.edu.au

More information about the Link mailing list