We have to balance efficiency with security. What we must avoid at all
costs is the entire network becoming dependant on one huge ubernode - or
a cluster of nodes all connected together and to few other nodes.
Locations must be reasonably well distributed across the network. I do
think that slow nodes should get a smaller slice of the keyspace, but
very large nodes should not necessarily get very large slices.

On Mon, Jun 05, 2006 at 03:25:20PM -0400, Colin Davis wrote:
> Good Afternoon,
> 
> As I understand it, the Freenet project has begun a Summer of Code  
> project to improve load balancing. I'm interested in seeing how  
> that's going, what the plans are, and how other people feel is the  
> best method of ensuring fairness across nodes.
> 
> I recall that about a month ago, Ian started a discussion on tit for  
> tat rewards to nodes, which rewarded nodes which were more useful to  
> the network. I'm curious how people feel best to implement such a  
> system, now that they've had to think on it, and while Michael Rogers  
> is working on his new code.
> 
> I'm also interested in ideas about how much to utilize highly  
> developed nodes- I understand that in general, it's better that  
> things are spread out to as many nodes as possible, but it seems  
> unoptimal not to max out the bandwidth and CPU of a supernode. If  
> say, MIT wants to donate a terrabit connection, with 500GB of HD  
> space, isn't it better that the machine is used?
> 
> As I understand things currently, .7 routing doesn't  the speed of  
> returning a block into consideration when looking at where to request  
> the next one. This probably isn't as efficient as it should be.. I  
> understand the need to establish the baselines first, but if we're  
> discussing how to make load balancing work, I imagine we want to keep  
> multiple ideas in mind.
> 
> I look forward to hearing how Michael is doing, and to helping to  
> discuss plans for the future.
> 
> -Colin
> 
> 
> 
> _______________________________________________
> Devl mailing list
> Devl at freenetproject.org
> http://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
> 

-- 
Matthew J Toseland - toad at amphibian.dyndns.org
Freenet Project Official Codemonkey - http://freenetproject.org/
ICTHUS - Nothing is impossible. Our Boss says so.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: 
<https://emu.freenetproject.org/pipermail/devl/attachments/20060606/ec93d908/attachment.pgp>

Reply via email to