[e2e] traffic dispersion and blocking probality

Alhussein Abouzeid hussein at ecse.rpi.edu
Tue Apr 2 07:19:46 PST 2002


Let's see if this is acceptable to you.

You send a number of stamped mail (i.e. you have paid for the stamps). You
wait for a response from your correspondent (say a month or so). If you get
one, you can send even more mail. Otherwise, you decide to wait for a couple
of years before resending (you also place stamps again, assuming you kept
copies of everything you sent).

That's a stable system indeed (the post offices will never "collapse").

But, would have preferred the post-office to accept my mail if they can get
it to the other end. I don't care if this requires a complicated
inter-post-office system design.

I'm not saying that we should go back to the ice age, where we used to
calculate blocking probabilities, but *some* (too many) interpretations of
"end-to-end" seem to be an argument against engineering.

Best,

Hussein.



----- Original Message -----
From: "Jon Crowcroft" <Jon.Crowcroft at cl.cam.ac.uk>
To: <jshen at cad.zju.edu.cn>
Cc: <irtf-rr at puck.nether.net>; <end2end-interest at postel.org>;
<OSPF at discuss.microsoft.com>; <Jon.Crowcroft at cl.cam.ac.uk>
Sent: Tuesday, April 02, 2002 1:56 AM
Subject: Re: [e2e] traffic dispersion and blocking probality


>
> In message <3CA80258.E126D151 at cad.zju.edu.cn>, Jing Shen typed:
>
>  >>I want to known if there has been some work done with routing path
decision and network blocking probality?
>  >>
>  >>Or, is there any paper with relationship between network blocking and
routing ?
>  >>
>  >>Each word will be highly appreciated.
>  >>
>
> well, if "blocking probability" == "not making useful progress due to
> congestion, and trying to avoid minor congestio ncollapse of even
> probing/trying", yes - there's Reslient Overlay Networking - uese
> altertante routes if the default ones are bust, and the Global Grid
> Forum people have built a simular system...
>
> ip routes around problems - if it doesnt, recurse - route around ip...
>
> j.
>




More information about the end2end-interest mailing list