Ok, I think I understand the model better. I see your point that there
is no locality to be exploited when there are too few peers. But I think
one interesting scenario is as follows especially considering that you
now have a powerful testbed to get controlled results. Fix the mapping
of peers to ISPs using some distribution (preferably empirically
derived) and then measure the effect of locality. I feel such an
experiment is still desirable since an administrator who would want to
deploy locality would likely not have the choice of using it only when
it is feasible (rather it is likely to be a simple blanket policy across
all torrents and for all ASes under his/her control).

> -----Original Message-----
> From: Stevens Le Blond [mailto:[EMAIL PROTECTED]
> Sent: Thursday, December 04, 2008 5:55 AM
> To: Satish Raghunath
> Cc: Arnaud Legout; Marshall Eubanks; [email protected]
> Subject: Re: [alto] Paper on "Pushing BitTorrent Locality to the
Limit"
> 
> 
> Dear Satish,
> 
> Thank you for your interest in our paper. I will  try to answer your
> questions, please do not hesitate to tell me if something is still
> unclear afterward.
> 

_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto

Reply via email to