As I know, it is common that one ip address can have different bandwidth due to dynamic assignment in reality. So I am afraid the cost of providing accurate bandwidth information just by address would be a little bit high. However this doesn't mean providing the location map has no sense for traffic optimization . Anyway, we hope the traffic can be localized as much as possible.
Best regards Syon DIng 2009/10/14 Reinaldo Penno <[email protected]>: > I'm aware of this issue but IMO this is not a specific ALTO issue and > therefore I do not think we should try to solve it. This is an ISP > Policy/Provisioning issue. > > The fast reuse of IP addresses have other impacts such as security, > accounting, subscriber limits, etc. The ISPs today are aware of this issue > since they are using or (trying to move) to a 'sticky' IP address approach > where the same device gets always the same address and/or make sure it takes > a 'safe' long time to recycle the same IP address to another customer. > > One could also argue that if the ISP is running the ALTO Service (or in > partnership with a Content Provider) it is in its best interest to > invalidate any IP/customer associated info whenever an IP address is dished > out by the DHCP/Radius Server. But again, I think this is a generic > provisioning problem ISPs face day to day. > > Thanks, > > Reinaldo > > On 10/13/09 6:52 AM, "Enrico Marocco" <[email protected]> > wrote: > >> Hi all, >> >> as I happened to discover recently, in at least two cases of fairly big >> ISPs IP addresses dynamically assigned to residential users are taken >> from pools regardless of the subscriber contracts. As a consequence, in >> such scenarios the same address could be quickly reassigned to users >> that, despite being topologically close, have very different provisioned >> bandwidth. So, while such reassignments are unlikely to happen to the >> peers participating in a swarm -- or at least are not going to have a >> significant statistical impact -- I can't figure any easy way to keep >> such information updated in relatively static maps some of the proposed >> solutions use (I can see how static maps AND dynamic queries can be used >> together though). >> >> Now, we had several discussions about the usefulness of information >> regarding provisioned bandwidth in peer selection without reaching >> consensus, and honestly don't know how common such an address assignment >> policy is. Thus I would be very interested in hearing what people think >> about that, whether it should be something reflected in the requirements >> and addressed by the solution, or simply just ignored. > > _______________________________________________ > alto mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/alto > _______________________________________________ alto mailing list [email protected] https://www.ietf.org/mailman/listinfo/alto
