Ungodly packet loss rates

Darin Wayrynen darin at good.net
Fri Oct 25 00:43:03 UTC 1996


> 
> 
>    >One can certainly make the argument that running large traffic
>    >flows through shared interconnects is bad engineering 
> 
>    Yes, and I think the operative word is "shared" as in "shared
>    access media".
> 
> Shared access to the media is not relevant.  Scalability of the shared
> media is.  Private interconnects are simply more scalable than shared
> interconnects.
> 

It makes sound engineering sense that networks pushing heavy amounts
of traffic between each other would like to distribute that load
around their networks, rather than try to take it all at the public
interconnects.

The private interconnects are not necessarily modeled after the public
exchange points, but are really ds3 or higher speed circuits
positioned between NSP POPs in areas were both networks have good
connectivity.

If NSPs continue to build out their Networks to support the public
NAPs, they would have to deal with higher and higher amounts of
concentrated traffic in those areas.  With private connects, NSPs can
distribute their load in less congested areas of their network.

Darin
-- 
                                       \////
                                      ( o o )
====================================oOO-(.)-OOo==========================
Darin Wayrynen, VP of Technology, (602) 303-9500 ext 3234, darin at good.net





More information about the NANOG mailing list