On Friday 05 September 2008 04:24, Daniel Cheng wrote:
> On Fri, Sep 5, 2008 at 7:29 AM, Matthew Toseland
> <toad at amphibian.dyndns.org> wrote:
> > Nextgens and others have been warning for some time that churn may be one 
of
> > the biggest challenges facing Freenet. How can we quantify this? 
Simulations
> > might be able to tell us something about the impact of churn on swapping,
> > data retention etc (when vive is done with the opennet/swapping 
project) ...
> > but we'll need some idea of what churn actually looks like. Is it the case
> > that 90% of our hits install Freenet, give it 5 minutes and then uninstall
> > it? (join/leave churn, in the Pitch Black terminology). What about 
low-uptime
> > nodes? How can we gather info on typical node uptimes? The reported 
uptimes
> > on a node's opennet peers are probably a rather biased sample, for 
instance.
> 
>   1) Low uptime nodes

How much of a problem are they? How do we find out?
> 
>   2) When a opennet node down for some minutes come back online,
>      they reseed from the seed nodes and connected to new peers
>      -- this make the node swap to another location.

Hopefully vive's results will enable us to turn off swapping on opennet, and 
therefore reduce location churn significantly.
> 
>  Last time we talk about this, we can't decide what to do on mixed
> dark/opennet node.
>  Maybe we can do this just on pure opennet node?
> 
> > Any ideas? It's probably worth simulating to see what kinds of churn 
affect
> > the network most badly, but we'll need to know what actually happens to 
see
> > where our energy should be best directed in terms of countermeasures and
> > tradeoffs. Thoughts???
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 827 bytes
Desc: not available
URL: 
<https://emu.freenetproject.org/pipermail/devl/attachments/20080905/529cd695/attachment.pgp>

Reply via email to