Re: Host caches and peer lifetime

This didn't get reply, but is a good question - we have been unable to
beat random selection of hosts, as high-uptime hosts are unlikely to
have available connection slots.  For a short while we would bias
slightly towards newer hosts with free leaf slots, but that started
creating some very weird "topological abominations" ;-)

On Jan 6, 2007 10:38 AM, Michael Rogers <[email protected]> wrote:
>
>
>
>
>
>
> Sorry if this has been covered before - do UDP host caches and
>  gwebcaches return a uniformly random selection of known peers? If so,
>  and if we can assume that peer lifetimes follow a distribution with a
>  heavier tail than the exponential distribution (ie old peers can be
>  expected to live longer than young peers), then would it make sense to
>  return old peers with higher probability than young peers?
>
>  Thanks,
>  Michael
>
>  http://people.cs.uchicago.edu/~matei/PAPERS/ic.pdf
>  http://2003.iwcw.org/papers/bustamante.pdf
>  http://www.imconf.net/imc-2005/papers/imc05efiles/stutzbach/stutzbach.pdf
>